Our complex world exists on a one dimensional spectrum ranging from beauty to chaos. When we get the best of our humanity, we create staggering accomplishments of art, sport, science and technology. In a single word: beauty. But when we fail to get the best of ourselves, we are just a massive system of interacting particles that feels the thermodynamic pull to entropy. At the bottom is horrifying chaos, the default state of the universe. Without constant energy input, all beauty decays. It is seriousness that drives a human system uphill against the pull to chaos and into new realms of beauty.

A democratic solution to our massive challenges would be the biggest cultural change a free society has ever accomplished. I want seriousness to be our new cultural battleflag because it encapsulates both risk and ambition in politically ambidextrous way.

This is a provocation. Because the formulation of seriousness is concise, just three intuitive dimensions with an estimation of activation energy, I use AI to “quantitatively” assess the seriousness of people and ideas. You can check it out yourself at IsXSerious.com

Shinya Kimura is serious about motorcycles. He hand-builds magnificent speed machines using his internal sense of beauty. 15 years ago I saw this two-minute video portrait of Shinya and it has permanently inspired me. Each of his machines is a singular work of art that takes six months, with collectors like Brad Pitt salivating for the next opportunity to buy. Shinya is not efficient, he is not market-optimized, and if Shinya did not exist, his designs would not emerge randomly. Yet he is playful and obviously in love with his work. He smiles wide while burying his head in the grease and steel: Shinya-san’s seriousness creates beauty.

I'm attracted to seriousness and serious people. By serious I don't mean grave or humorless (Shinya clearly has fun), but rather an earnest, sustained commitment to creating something excellent. I’ve come to see the world as a spectrum between chaos and beauty, and seriousness is the force that resists the entropic pull towards chaos while taking us into new bounds of what’s possible.

The more I pondered what differentiates seriousness from unseriousness (or, worse, anti-seriousness), the more I saw that we broadly lack principles and orientation for what generates incredible work. We do, however, have a surplus of language to tear down, critique and complain1. This is the intellectual trap of feeling productive, perpetuating the fallacy that it takes the same amount of talent to destroy as it takes to create2. Making “not an ugly statue” will never produce the Venus De Milo no matter how many times you try. Pointing to China alone won’t ever make American science policy durably better3. So when I think about the science+tech transformation we urgently need, I’m done with looking sideways (“Xi!”) and backward (“Vannevar Bush!”) and suggest we put our energy to framing a self-contained set of principles that takes us into the future.

We’ll know that we have cultural alignment on creative principles when have a shared rallying cry again. But right now, we don’t have anything good enough. Consider our options:
“Efficiency!” is an essential component to successful execution but is not a vision that guides us into the future4.
“Diversity!” gave us false prophets but will never create a marvel like the Sistine Chapel.
“Get rich!” motivates individuals, but alone it leads to opportunism like stock fiddling, influencer peddling or charging $2500 for an Advil in an emergency room.
“Prestige!” is similar to riches, everyone wants to feel fancy but that gets easily twisted.
“Progress! Abundance!” is incomplete because it intentionally sidesteps the darker sides of human nature.
“America!” currently activates ~60% of the country at a time until we rediscover our national espirit de corps5.

“Seriousness!” is a politically ambidextrous battleflag that acknowledges the dark sides of nature while celebrating beauty and progress.

It is a problem that science policy people generally avoid impolite topics. Many will read Derek Thompson’s Abundance, Ben Reinhardt’s SpecTech Newsletter and Jason Crawford’s Techno-Humanist Manifesto. They are indeed good and they rally the base of people who care about science. But too few have read The Sovereign Individual (what does money mean when a state loses the monopoly of violence?) and Gulag Archipelago (how did the Soviet prisons kill 10x more than the Holocaust and why did the people accept that?). Talking progress without talking primal needs is like watching an adventure movie but skipping the scary parts: there’s some cool scenes but it’s not a film.

We’ve grown so separated from each other and from consequences that major problems have brewed. Consider three examples that should motivate us to get serious:

• We sleepwalked into a shortage of materials to build the future: The horrific Aug 2025 USGS Report on Critical Mineral Supply Chain Risks shows the USA no longer has the metals to build things ourselves. Within 24 hours of reading the report, my colleague and I created the Minerals Explorer, to understand the bottlenecks and technical challenges to WHY the USA has so little capacity. The USA risks $64B in damages if we can’t get ~10 tons of rhodium in a place other than South Africa and faces a halted electromechanical industry if we can’t onshore smelting (eg, US has just four aluminum smelters) or alternatives to 40-stage solvent extractions for rare earth elements. You can then go through government reports on Thorium, Potash, Gallium and others to see “Government Stockpile: None” while wondering what flavor of over-financialization let this happen.

• Ubiquitous Disease compounds our already-impossible demographics: I estimate only ~50% babies born in 2025 will be able-bodied and sufficiently psychologically resilient adults to support themselves in the workforce. Compare this to the roughly 75–80% work-readiness rate of Americans born in the mid-1990s. I base this estimate on sharp rises in autism, childhood obesity, developmental and mental health disorders, and screen-related neuronal dysregulation. On top of this, consider the babies born in 2025 will have a ratio of ~2 workers to 1 retiree on security (it was 42:1 in 1945, 2.7:1 today, ~2:1 by 2045, 11 states already have more seniors than children). So a future adult born today will be working to support somebody himself, his family, and either a random person his own age or a GenX retiree.

• We have a budget that constitutes a national security risk: America borrows $1.9 trillion a year from countries it presently considers a threat to national security. American debt, once the global safe haven of capital, has now been downgraded by all three ratings agencies, an indicator of decreasing confidence of future capacity to ever pay back the debt. It would not take much6 from a maligned set of actors to tank American debt and render the dollar worthless.

A democratic solution to our massive challenges would be the biggest cultural change a free society has ever accomplished. Yet we have to figure it out. The saddest failure mode of the USA is a country that makes nothing, imports everything and prints increasingly useless money to run a welfare state to nowhere.

Embracing seriousness means shifting the elite student’s dream job from well-manicured analyst to hands-on, risk-tolerant industrialist. It means popping the balloon of vague prestige, ignoring the student protest du jour and celebrating those who get into the dirt and dig for gold themselves. It means appreciating that farmers are actually very smart, expert machinists are in high demand for a reason, great inventors/scientists are not a commodity and that those in the physical defense roles (military and police) are worthy of your gratitude. It means active measures to make all children healthier and fearlessly increasing standards to compete on any field. We must identify and appreciate seriousness while jettisoning our patience with unserious or anti-serious behavior.

Loosely, you know what I mean by “serious,” but it’s worth defining properly.

Using foundations of physics and lessons from biology, I see three dimensions to the seriousness of a system: Energy, Infrastructure and Order. Together, these describe the balance of a system that either grows or succumbs to the entropic pull down to chaos. Each act of seriousness is a substantial effort that overcomes an activation energy toward a stable, more serious state.

So while a non-living ball rolls down the energy hill into chaos, Shinya Kimura gasses up his tank, harnesses the combustion energy through his machine and climbs the hill of human accomplishment.

This is a long essay: I’m going to do brief dive into physics, share the seriousness checklist, then measure the seriousness of well-known people and efforts.

Deriving seriousness from a stasis with chaos

Paint chips, gears wear, roofs leak, plants die. With no energy input, every system ultimately decays to nothing. How fast that decay happens is a property of that system: it’s obvious that a rose will decay much faster than a cube of tungsten, but why? The balance of needed energy and system structure is a thermodynamics question and the unexpectedly useful foundation of seriousness.

The Second Law of Thermodynamics says that, in a closed system, entropy only increases. Hot coffee cools, gradients flatten, order decays. So at first glance, life seems to be an exception. Bacteria multiply, flowers bloom. But a living organism is an open system, energy from outside comes in and the organisms converts that energy into molecules and gradients for growth. If any part of the exquisite machinery goes wrong, the organism succumbs to chaos and its molecules are recycled.

We can write an equation to show the necessary flux of energy to keep a system (whether a cell or a city) balancing between growth and decay. I took the Second Law, unpacked definitions and considered the time derivative:

On the left side is the resource flux: some energy flux Ė (energy per unit time, aka Power) and some capture percentage v. A rosebush captures the Ė from the sun at some efficiency v (~1% for photosynthesis), and as the plant grows, more surface area is exposed to the sun, increasing Ė.

On the right side is the “entropic pull”: The more complex a system, the more possible states in the system which means the more ways it can fall into chaos. The intricacy of the growing system of N parts with α interactions between parts describes the many possible states. The dark truth of the universe is that only a tiny fraction of the possible states are actually good (see the Anna Karenina Principle).

The intuition of seriousness is simple: if a system can’t bring in enough energy to manage its complexity, the system falls into progressively worse states before collapse.

So a serious person contributes to the growth of her system in the face of chaos. To do so, she must accomplish at least one of the following:

• Increase the total energy, Ė: Energy is Life7 (’s refrain, see below)

• Increase the infrastructure, v. Improve the system’s ability to use resources.

• Manage the internal order, α. Keep the system vibrant but not anarchic.

Doomberg

It's Time to Get Serious About Energy

“And so, I hope President Putin will help us to stay on track with respect to what we need to do for the climate.” – John Kerry, February 23, 2022…

Read more

4 years ago · 1632 likes · 135 comments · Doomberg

Serious things requires activation energy

Now we have Ė,v and α to determine if an effort is pointing in a serious direction, which is necessary but not sufficient. There has to be enough substance to the effort for it to matter. At the atomic scale we would call the threshold of input the activation energy.

In chemistry, energy breaks old bonds to then form new ones. The amount of needed activation energy to break these bonds, and the difference in stored energy in the new bonds, determines if the reaction would just happen spontaneously by itself. Every serious act at the human scale—starting a company, raising a family, growing a city— requires it’s analogous concentrated burst. To be serious is to rally sufficient resources to make it to the next stable configuration (aka, a milestone).

The seriousness checklist

Seriousness can only be judged within a context. For example, “Was Mr. Rogers serious in the context of American culture?” or “Is Dan serious about building a time machine?”

Condensing what what we’re built so far, this is the seriousness checklist:

• Direction: Does the effort push meaningfully uphill along one or more axes of improvement — energy harnessing (Ė), internal coordination or order (α), or adaptive capacity (v)?

• Magnitude: Does the effort possess or mobilize sufficient activation energy to reach a new stable state? Serious actions don’t just oscillate; they cross thresholds. Is there enough force, focus, or persistence to make the transition real?

• Stakes: Is there a clear vector of consequence between what is feared or rejected (the decay state) and what is desired (the beauty state)? In other words, is there something real at risk — a meaningful difference between success and failure?

• Fungibility: Would this progress occur without the agent’s directed effort?
  Seriousness is inversely proportional to inevitability. If the outcome would happen anyway, the action carries little Seriousness.

Quantifying Seriousness

I expanded the serious checklist into a full LLM prompt, you can see the full prompt at IsXSerious.com/prompt.html. Now we can explore how good AI is at assessing a human value.

While this might feel like we’re getting adjacent to things like the Social Credit Score in China, I hope the open source nature of seriousness makes it more democratic and transparent.

What you’ll see in this section is something like (Ė=-0.7, v= -0.6, α=-0.65). The AI captures both the direction (positive is serious, negative is anti-serious) and also magnitude, which is normalized proportional to the activation energy.

We’re projecting human ideas and history onto 3D space. When all the numbers are positive and large, it’s a Serious Effort. If all the numbers are small, then it’s unserious. But if the vector is large and it’s negative, then it’s anti-serious. Back to the Anna Karenina Principle, there’s much more room for anti-seriousness than seriousness.

For example, the offshoring of American manufacturing to China was profoundly unserious in the context of the United States (Ė=-0.7, v= -0.6, α=-0.65). The consultants and bankers that did the offshoring work were unserious and fungible (any person could have been the one to see that margins were higher and short-terms risks were lower) regardless how much money they made.

On the other hand, the offshoring of American manufacturing to China was profoundly serious in the context of China (Ė=0.85, v=0.9,α=0.7). The Chinese leadership played the game of volume, not margins8, which is frankly a remarkable effort of vision and leadership. It changed the energy landscape of global manufacturing, making it inevitable that the bulk of the world’s building would drift down the energy gradient toward China.

To show the AI’s rationale, let’s explore two political movements in the US. Both below are the values and the rationale right from the Anthropic API call:

• “White Fragility” in the context of United States (Ė=-0.06, v=-0.08,α=-0.12)9: “[White Fragility] demonstrated negative impacts across all three dimensions. It consumes institutional resources while producing limited measurable improvements in racial equity (E). The implementation has reduced organizational efficiency by creating communication barriers and increasing workplace tension (v). Most significantly, it has increased social disorder by heightening racial anxiety, essentializing racial categories, and reducing cross-racial communication (α). While the goal of addressing racism is serious, this particular conceptual framework and its implementation have produced effects counter to system cohesion and productivity.”

• “Project 2025” in the context of United States (Ė=-0.45, v=-0.56,α=-0.68): “Project 2025 qualifies as anti-serious because it would likely reduce system-level resources (E), damage institutional conversion efficiency (v), and increase internal disorder (alpha). The project would redirect resources away from established programs, potentially creating net resource loss during transition. Its proposed dismantling of institutional knowledge and civil service expertise would reduce the government’s ability to effectively convert resources into services. The polarizing nature of its implementation would significantly increase social entropy, pushing the system toward greater disorder and factional conflict rather than coordination.”

Moving to higher throughput, here are 40 contemporary American politicians (left) and all American Presidents (right). Note: John Lewis is listed as the most serious on the left, and there is a tie between George Washington and Abraham Lincoln for the most serious president. (View the full raw data)

And exploring a different way to visualize the data, below is a different visualization of 20 of the biggest technology ideas of the past two decades. (View the full raw data)

There are many other sets of people and ideas we can measure with this. I’m curious what your ideas would be, and how others would improve the prompt.

Conclusion: Heroes and Heroic Efforts

Seriousness can be the positive battleflag for us to rally around. It directly values efforts that make a system be holistically better and there is no way we will solve our problems without being more collectively serious than we are today. In my heart, we have it in us to one day see the same view as NASA’s Cassini, looking back at Earth from Saturn’s shadow. The chaos at our doorstep is what holds us back from the beautiful futures we dream about.

I want to hear “Oh yeah, she’s a serious person!” as a recommendation for a job and “Ooh that is is a serious idea” as an endorsement for an investment. I want the political right to be excited about building serious science institutions and I want the political left to see the anti-serious foundations of their last decade. I’d like a student to get that classic question of what they want to be when grow up and for them to give the irreverent answer “Pssssh, I just want to be serious to my community!”

I hope every reader asks themselves “Am I serious person? If not, how do I become serious?” And if you don’t yet feel serious, that’s OK. I can give you a cheat code: find heroes.

The essence of being serious is having both direction and capacity for great work that improves the world around you. Heroes embody this very notion. The salty Youtube channel Critical Drinker has a slightly offensive yet surprisingly beautiful tirade about the importance of heroes using the lens of comic books. Heroes may just be humans, but they are high agency people whose serious efforts change how we see the world. Learn from them.

I’m blessed to have my family as heroes, and friends and mentors that are heroes, and I have Shinya Kimura as a hero, too. His impact is clear in this work. I didn’t need to handcraft this essay for six months just like he hammers his motorcycles. But, like Shinya showed me 15 years ago, sometimes a bit of obsession and weirdness and raw effort can push through the activation barrier and put some new beauty into this world.

Epilogue:

If you made it this far, how about some “Get Serious” Trading Cards?

(Now taking requests!)

Appendix: A brief exploration of α

Which would take more energy to manage: a classroom of fifth-graders or a gym of professional fighters? The answer is obvious, and it points to the meaning of α.

Ten-year-olds are chaotic by nature, colliding with everything. Professional fighters, though far more capable of destruction, are both internally and externally ordered. They interact precisely. They conserve energy. Abstracting this idea: If you consider these two social scenarios as networks with N nodes (people), then let N^α be the approximate number of edges. The classroom is α ≈ 2 (everybody interacting with everybody), the martial arts gym is α ≈ 1 (roughly one partner per person).

It turns out some network scientists were actively exploring similar questions a decade ago. Luis Bettencourt’s Origins of Scaling in Cities (Science 2013) estimated the scaling factor α ≈ 0.85 for the energy usage of urban networks. This is less than theory predicted but points to the efficiencies of infrastructure offset social complexity. It also captures the real world constraints: if I don’t physically live by somebody, I probably won’t have much interaction with them in my city. Online, however, the numbers are much higher: between 1.9 for Spanish Wikipedia (2013 Nature) and 2.5 for unmoderated Russian News Sites (2017 Applied Network Science). Intuitively, the more unbounded the environment, the more chaotic the interaction space.

We therefore hit the tradeoff as old as human history: Blindly minimizing α takes you to an Orwellian prison or the Gulag Archipelago. But the opposite extreme is just as scary: high α is a chaotic, mob-ruled, violent and fractionated bag of humanity. The ancient wisdom of religions and philosophy often serve to temper the α parameter appropriately, balancing individuality and social buy-in.

My goal is not to be prescriptive for α values. But I must point out that the real world is blending with the virtual world, which means the α parameter is going to drift higher. A mean comment on a post is nothing, but an act of violence from a stranger could be everything.

Ask any Nobel Prize historian, and they’ll tell you: sometimes, all it takes is a few visionary donors to spark a scientific revolution. Think of how Warren Weaver’s well-timed grants helped underwrite discoveries in molecular biology—ultimately leading to 19 Nobel Prizes in that field. History is full of similar stories, from the Rockefeller Foundation’s crusade against hookworm to modern philanthropic drives eradicating polio. Now, one stealth team has quietly emerged, setting out to beat them all in sheer audacity: The Focused Research Organization (FRO) for Time Travel.

Yes, you read that correctly. This newly minted group has a tidy $50 million plan to manipulate spacetime—in a controlled research environment—for the good of humankind. Forget incremental progress or “sandbox” pilots: these folks believe targeted philanthropic funding can actually re-engineer our timeline.

Why Time Travel?

Because preventing the next genocide or pandemic is more than worth a wild bet. It is a moral travesty to NOT try to recover the 8+ million lives lost in COVID and the immeasurable toll of 20th-century atrocities. The philanthropic logic is disarmingly simple: if there’s even a remote chance to preempt global disasters before they happen, that chance demands serious inquiry.

High-risk, high-reward: It’s precisely the sort of moonshot that philanthropic capital can champion when no government agency or Fortune 500 R&D budget is likely to step up. After all, “time travel” has historically been relegated to science fiction. But so were nuclear reactions—until the Manhattan Project. So were quantum computers—until Google demonstrated quantum supremacy. The FRO for Time Travel aims to be the next watershed moment of disruptive science.

The $50M Blueprint

1. Building a Time Distortion Device (TDD)

Leveraging advanced ring-laser arrays and gravitational frame-dragging theory, the FRO’s team wants to create controlled “closed timelike curves” (CTCs). In practical terms, it means swirling laser beams and rotating superconducting matter at speeds that theoretically bend spacetime enough for signals—maybe even small objects—to loop backward by a few nanoseconds or more.

2. Generating Exotic Matter (Casimir Effect at Scale)

To stabilize wormhole-like structures, you need “negative energy.” Enter the Casimir effect—a quantum phenomenon that can exhibit negative energy densities between parallel plates. Usually, it’s microscopic. The FRO’s plan? Scale it up by orders of magnitude, possibly harnessing the weird properties of vacuum fluctuations. Yes, that’s like saying you’d tap into the quantum vacuum to create space-bending “exotic matter” on demand.

3. The Jupiter-Sized Elephant in the Room

The critics say you’d need to convert something on the order of Jupiter’s entire mass into energy to produce the necessary spacetime curvature for a stable, human-scale time loop. The FRO’s official stance? “Good.” We argue that historical breakthroughs happened because visionaries refused to be boxed in by immediate constraints. With well-deployed philanthropic dollars, we might find novel energy amplification or resonance methods that dramatically reduce the “Jupiter factor.”

Sound insane? In the spirit of the best TED talks, the FRO gently reminds us: “Impossible was once said of heavier-than-air flight.”

Technical Foundations: The Physics of Bending Time

The backbone of this initiative lies in General Relativity (GR), where Einstein’s field equations imply that mass-energy warps spacetime.

Certain solutions, such as the Tipler Cylinder and Gödel Universe, accommodate closed timelike curves (CTCs)—theoretical loops through spacetime permitting backward travel (Gödel, 1949 and Tipler, 1974). However, these classical solutions typically demand either unbounded mass-energy density or cosmological rotation. The FRO’s approach borrows from Morris–Thorne Wormholes (Morris, Thorne and Yurtsever, 1988), seeking to generate and stabilize a compact region of spacetime curvature by integrating exotic matter—which, in semiclassical gravity, can exhibit negative energy densities (e.g., via the Casimir effect, first quantified by (Casimir,1948 and extended by Ford, 2010). Specifically, we aim to place high-intensity ring lasers inside a resonant, superconducting structure to simulate the frame-dragging normally seen in massive rotating systems. In simplified form, one can approximate the local spacetime metric with an off-diagonal term that tilts the local light cones, creating a CTC if Ωr surpasses a critical threshold (Tipler, 1974). Yet bridging the gap between microscopic “test” loops and macroscale CTCs demands unprecedented energy densities—on the order of converting a Jupiter mass to energy (10) unless offset by resonant amplification.

Parallel to this, the team’s Casimir experiment leverages quantum vacuum fluctuations between parallel plates: the vacuum stress leads to an effective negative, described for perfectly conducting plates in 1D by (Casimir,1948) and revisited in (Ford, 2010).

In practice, we plan to scale up plate arrays using advanced nanomaterials and waveguide architectures, thereby enhancing the local negative pressure zone. By coupling this negative region to a rotating or circulating laser field, the group aims to realize a partial violation of classical energy conditions (the “weak energy condition”)—a key theoretical requirement for stable wormhole geometries (Morris, Thorne and Yurtsever, 1988 and Ford, 2010). Though daunting, it’s not disallowed by known physics. We argue that philanthropic-scale resources, followed by aggressive government investment, can push the experimental envelope beyond current one-off lab tests—potentially achieving a self-sustaining negative energy cavity that warps light cones enough to detect genuine backward time shifts.

Why Could This Actually Work?

• General Relativity Doesn’t Forbid It: Certain solutions of Einstein’s field equations—like the Tipler cylinder or traversable wormholes—do permit loops back in time. Their existence is mathematically valid (though physically extreme).

• Quantum Loopholes: Negative energy is real (Casimir effect), so the door to exotic spacetime configurations is open, if only a crack.

• Philanthropic Leverage: Large-scale private funding can unify top physicists, engineers, and a state-of-the-art lab under one roof, free from the usual short grant cycles.

In other words, the puzzle pieces exist in theoretical physics. The FRO wants to see if $50 million can knit them together.

Potential Upsides

1. Prevent Future Crises: Imagine quarantining a new pathogen before it spreads or halting genocides at the planning stage.

2. Revolutionize Energy & Materials: Even the partial breakthroughs in “negative energy” or high-intensity laser arrays could trickle down into spin-off tech.

3. Historically Unprecedented ROI: As one philanthropic adviser notes, “If time travel works, there’s literally no bigger win.”

Put simply, to NOT fund this FRO is to endorse the atrocities of human history.

Risks & Rebuttals

1. Paradoxes Galore: The dreaded “kill-your-own-grandfather” scenario. The FRO’s solution? Strict internal protocols for “consistent” time loops—no self-contradictory interventions.

2. Safety Concerns: High-energy lasers, potential vacuum instabilities—these are not backyard experiments. The FRO compares their risk management approach to the Large Hadron Collider’s rigorous safety reviews.

3. Ethical Quagmires: If you could rewrite history, who decides which events to erase or preserve? The organization claims they’ll form an Ethics Advisory Board to handle policy guidelines and moral oversight.

From Stealth to Spotlight

Until recently, the FRO for Time Travel was operating in stealth mode—quietly recruiting top theorists, experimental physicists, and even interdisciplinary philosophers. Our unveiling signals a call for philanthropic allies who want a seat at the table—preferably those with an appetite for high-stakes scientific leaps.

They’re not aiming for incremental academic papers: They want to make a functioning prototype that can show any tangible backward time shift—be it microseconds or days. The rest, they say, is scaling.

Your Move, Funders

In an era where big names chase AI, quantum computing, or longevity research, this FRO stands out by targeting a dimension we scarcely thought we could alter: time itself. They argue that, like the major philanthropic bets of the past, success here would transform everything—the ultimate “disruptive innovation.”

Will $50 million be enough? By some measures, it’s a drop in the ocean. Yet this figure is often all it takes for a small group of dedicated experts to crash through a theoretical barrier. Whether they fail fast or break the universe (kidding…mostly), the attempt itself could spawn new insights across physics and technology.

Next Steps

• Read our 20-page proposal: Even if you doubt time travel is feasible, the spin-off advances in lasers, superconductors, and quantum phenomena may prove revolutionary.

• Spread the Word: Share this unveiling. The FRO is no longer a secret, and they’re seeking bold partners.

• Ask the Hard Questions: Paradoxes, ethics, energy demands—address them head-on. That’s what the FRO claims to be doing with its top-tier team.

References & Further Reading

1. Hawking, S.W. (1992). “Chronology protection conjecture.” Physical Review D, 46(2), 603–611.

2. Morris, M.S., Thorne, K.S., & Yurtsever, U. (1988). “Wormholes, time machines, and the weak energy condition.” Physical Review Letters, 61(13), 1446–1449.

3. Tipler, F.J. (1974). “Rotating Cylinders and the Possibility of Global Causality Violation.” Physical Review D, 9(8), 2203–2206.

4. Casimir, H.B.G. (1948). “On the Attraction Between Two Perfectly Conducting Plates.” Proceedings of the Koninklijke Nederlandse Akademie van Wetenschappen B, 51, 793–795.

5. Ford, L.H. (2010). “Negative Energy Densities in Quantum Field Theory.” International Journal of Modern Physics A, 25(14), 2355–2364.

OK, what did I just read?

This whole thing, blog and 20-page paper, was generated with ChatGPT’s “Deep Research” Tool. This is the entire chat, including the writing of this blog post. It took me an hour.

All in: 12 minutes for the O1 Model to do the research, 20 minutes of back and forth on the blog with GPT, 20 minutes to fix equations in substack. Add in a few minutes for copy/paste into Midjourney and some laughing breaks.

The purpose of this was to be a demonstration of the post-AI world we now live in. If we judge things by form alone, at high-level, cool-sounding ideas, then we’re now able to convince ourselves of anything. Funders especially have to understand that fancy ideas are free now. But, in the best outcome, we’re entering a golden era for those with great ideas, great taste and great capacity for execution.

In my own head, I think a lot about education, about building teams to do great work in science, and how to partner best with capital allocators. We all need freedom to operate and a pragmatism that some legacy structures might not apply any longer. If we engage with AI correctly, AI’s capacity becomes the floor of human capacity, empowering an incredible future of human creativity and output.

I hope you enjoyed :)

The night before the election, I shared who I was voting for. I riffed on the Harris campaign’s refrain of “saving democracy,” arguing that saving democracy is more about how we interact rather than who we vote for. It’s more important to be human and connected than ever before. Still, my choice of Trump shocked a lot of people. I received many private messages in response, the vast majority were positive or at least curious. The most constructive conversations normally started with a “I deeply disagree” but ended on an affectionate note. If the conversation really went well, the last thing people would ask me is “Dan, what makes you optimistic about this new administration?” Here is my answer.

Four days after the election, I’m walking through the Weir Hill nature reserve and found the metaphor for American Science in 2025. Weir Hill “restores their habitat with fire.”

In contrast to forest management to prevent wildfire, which have led to static ecosystems becoming tinderboxes, Weir Hill takes the opposite approach to cultivate dynamic, young ecosystems."Without active management this area would succeed to mature forest, and this rare habitat and wildlife that depends on it would disappear." In the northeast, the result of fire is transient blooms of wild indigo which then host beautiful species like the Frosted Elfin Butterfly, endangered because of the lack of natural burns. In other forests, it takes fire to releases seeds from the pinecones to grow into the next generation.

We all know a forest fire is coming to science. I believe we can make this a good thing.

Some of the old growth of our academic ecosystem is beautiful and essential, yet just like in a badly managed forest, the broader ecosystem got stuck in an artificial stasis. If science progresses one funeral at a time, then maybe nation-scale science progresses one DOGE at a time.

Like everyone else, I’m getting my own bearings and developing my mental models for a volatile few years ahead. Before getting to science problems and possible solutions, it’s worth being grateful for an uncontested election and appreciating the path not taken.

Subscribe now

Let’s be thankful that we got one of the two best outcomes in the election

Trump getting elected is not the end of the world, but Trump being killed on the campaign trail would have been the end of our world as we know it. I don’t think this is appreciated widely enough. We missed civil disaster by inches and all we got was the most iconic political photo in my lifetime.

The two best outcomes were a landslide victory for Trump or a landslide victory for Harris. Anything else would have been dragged into court, raised claims of fraud, created more online hysteria and decreased confidence in the democratic system. Worse, there would have been reputational damage done to the US at a time of increasing multipolarity, which would have hurt the investability1 and leadership capacity of the US.

Peter Thiel had a memorable on-stage comment over the Summer when he was asked if he’d engage with the politics in this election. Paraphrasing, “I’m going to sit this one out because I have nothing to add. If Trump wins, it’ll be a landslide. If it’s close, Harris will win by cheating.” After some gasps, half the audience booed, half the audience cheered. That’s 2024 politics encapsulated in a 30-second exchange.

Thiel was right. Trump swept all seven swing states and won the popular vote by 2,475,506 voters. Riding Trump’s popularity and discontent with the Democratic incumbents across the board, the House, Senate and Governorships also went red. This is clear victory, likely the most dominant and odds-beating election in my life, but what matters most is that it’s done and we can move on.

The big story is individuals over institutions

A few years ago, following a prompt by Tyler Cowen, I wrote the essay “Is the world getting harder to predict?” I concluded that yes, the world is getting harder to predict because although large organizations are easy to model, individuals are not, and it’s individuals that are increasing in power. Worse, larger institutions like universities, the NIH and national political parties (DNC and RNC) are both increasingly predictable AND dysfunctional.

Kamala Harris was always just a representative for the Democratic Party. She refused to do anything that define herself as an individual beyond her institution. So this election was a head-to-head of two very different teams: On the blue team, the Democratic National Convention, a billion dollars, legacy media, and academia. On the red team, Trump, Vance, Elon, RFK, Tulsi, Rogan, Thiel, Tucker, Vivek, Lonsdale and their respective empires. Team Blue ran the predictable same plays as the previous two campaigns. Team Red did weird things and spoke their minds. Team Red won bigly.

The individuals vs institutions dynamic is going to play out first in government agencies then in science. The biggest trees —the biggest, oldest government science institutions— are going to burn the hottest. But on the other side, we may have an innovation ecosystem that empowers, rather than stifles, the individual.

Talk to professors today, and they’ll tell you the situation today is untenable: flat science funding, sky-rocketing costs and an ever-lengthening path to career stability. Creative destruction is necessary, and sudden jolt is needed to break the inertia and get us on a better path. Without such disruption, we risk following the troubling trajectory of similar countries.

A Harris presidency probably would have mirrored the inertia of other countries

It’s fair to say that a Harris presidency would have looked like our more liberal, institution-centric allies. Canada and Germany, for example, look smart but they are not doing well. We don’t want to be like those countries right now. As I said in the last blog post, “Maslow is real: Climate policy means nothing if we have a failed currency and unsafe place to do the work.” Consider:

Climate sits inside of science,
Science sits inside of an economy,
An economy sits inside of a functional and safe country.

Put coldly, if there is ever a World War III, nobody will care about CO2 levels in 2080. A vibrant science ecosystem can only thrive with sufficient funding and stability to look ahead.

Negative GDP growth may be less hyperbolic than nuclear war but it has the same impact of squashing the long-term, ambitious spirit that gets people in science (especially climate) in the first place. If the people are unhappy, or if the government has to shut off the power to universities to curb rampant inflation (as is happening in Argentina), science is the first to go.

Canada is a yellow warning light: Canadian GDP has grown a paltry 4% in the past decade (the US has grown 47%) and now the cuts are coming to Canadian science: among others, ~20% off the Canadian version of the SBIR despite the government’s own figure that every $1 into that program produces $3.10 of GDP and $1.40 of tax revenue. How bad are things if they are cutting programs that pay for themselves?

Canada can fix their issues but Germany seems to be a full-on goat rodeo. It’s likely going into a depression due to its inability to adjust course from institutional inertia. And this one feels personal to me.

In 2018, new President Trump went up to the UN podium and said “Germany will become completely dependent on Russian oil if it does not immediately change course.” The Germans mocked him and this moment went viral in my networks along the lines of “look how embarrassing our President is!” I remember watching this moment in lab and feeling the emotions clearly: we in biotech academia all looked to Germany as the country of smart people. But six years later, Trump was right and those German leaders have been horrifically wrong.

It would have been bad enough if Germans were “just” stuck on Russian energy since the war started as Trump warned. But astonishingly, the progressive Germans still proudly dismantled their nuclear energy capacity, an own-goal on 25% of their baseload power, in the middle of an energy crisis due to a kinetic war next door. This was institutional arrogance mixed with corruption: only 25% of the country actually wanted this. The dominos fell: Energy costs tripled, government subsidy programs vanished with tax revenues, green technology factory plans got scrapped because nothing could be made profitably in that environment, the government hit their Debt Brake. In a Murphian comedy, there was no wind in Germany’s sails the week Trump got elected: wind power vanished leading to blackouts despite the €500-€750 Billion Energiewende electrical grid overhaul. So Germany’s coalition government collapsed then they dropped out of COP on November 7. No matter how polished, no matter how passionate the climate rhetoric, Germany’s capacity for global impact is minimal until massive changes are made.

It’s more than just Germany: Europe’s trends are so bad that the EU commissioned former Italian Prime Minister Mario Draghi to analyze the stagnation. Noah Smith does a pretty good walkthrough of the 328-page report, and it’s bleak and contains themes of institutional bloat. Back in the USA, the domestic Overton Window is shifting rapidly: Nate Silver just roasted the Expert Class in the US and Matt Stoller’s anti-monopoly blog aligns more and more with Trump’s appointees. I would not have expected these outcomes in 2015.

So I think another Democrat Institutional President doing more of the same would have kept us in the boat with Germany, the UK and Canada. There would have been some superficial changes to address some problems in science. And maybe some sweeping Executive Orders from President Harris, maybe some billion-dollar programs. But I think they all would have been small perturbations around the same negative trend. As my wife once said, “sometimes a big change is easier than a small change” and I think only a big change gets us onto a positive, long-term viable path.

Here are nine problems of science in 2024

Everybody agrees that our science must be excellent, and everyone agrees it could be a lot better than the status quo. I like Eric Schmidt’s term Innovation Power — “the ability to invent, adopt, and adapt new technologies” — when he argued for Innovation’s importance to a country’s hard and soft power. But before we can start to pitch solutions, we have to start framing problems with our current innovation ecosystem:

1. The USA sends ~$54B/year into universities for research despite 90% drop of return on investment (evaluated on disruptiveness) since 1980 and majority disapproval of universities. Less than half the country trusts universities, likely because of excessive politicization, and approval numbers continue to drop.

2. At least one third of degree tuition, $130 Billion per year, is wasted on degrees that are worth less than their price tag. What would you study at Boston College if you were paying $92k per year at a school with no merit-based scholarship?

3. Elite academia has become an exclusive and brittle monoculture. DEI and Covid origins are two well-known examples of a tightrope of acceptable discourse, marked by aggressive self-policing. Two good (long) essays of the status quo are from Claremont McKenna College and from Stanford. Administrators continue to hockeystick - MIT hired 6 assistant deans of DEI in 2021. Prestige is an unstable equilibrium: once you’re no longer the best, the fall is steep.

4. Universities are progressively failing to recruit and maintain top talent. Top academics are leaving for 10x salaries at industry labs. Chinese nationals are going back to China at an exponential rate, top graduates just go straight to industry. The starting salary as a professor at Harvard is $125,000, which is barely enough to raise a family in the area, and comparable in pay to a lower-stress administrator job.

5. The awfulness of postdoc and early professor life scares talent away from science. Only ~20% of postdocs get faculty jobs. Then Professors spend 40% of their time applying to grants and are stuck in their lanes despite wanting to branch out (Fast Grants Survey). This is why many PhDs drop out to work in data science or parlay their grad school acceptance into a financial services job.

6. Venture Capital outcompetes the government for superstar talent for advanced research funding. When DARPA were invented in the 1950s, Janet Yellen (today’s Treasury Secretary) would have earned more than David Solomon (CEO of Goldman Sachs). Today, David makes 100x Janet’s salary. You can’t blame young talent for wanting to be David, not Janet. Or, in technology, young stars want to be Sheryl Sandberg, not Evelyn Wang (ARPA-E Director).

7. The US has created ONE elite university since 1955. There have been good state schools, but the energy and investment to create a great school, like a Stanford (1885) or an MIT (1861) or a Harvey Mudd (1955) or Olin (1997) just isn’t present right now.

8. The American innovator is stuck with a duopoly of universities or startups to develop her technology. As I explored it the Problem-Centric Founder, startups and universities do *not* span the whole space of needed work. A defeatist “There Is No Alternative” mentality to funding scalable research at places other than universities.

9. Philanthropy has been leading the experiments with scientific ecosystems but it’s a bridge to nowhere until the government learns to work with these new organizations. Ben Reinhardt’s “Fund Organizations, not Projects” 2022 policy piece articulates and addresses a pain which all philanthropically-backed organizations feel. Cultivating the founder phenotype in science can be an unlock for science, but that requires building an ecosystem with full organization lifecycles from seed checks to acquisitions.

Here are three ideas for science

The first phase of the Trump II Presidency is probably going to be focused on cutting rather than building. Still, there’s a natural appetite for seeding positive action in the first 100 days of a presidency. For innovators with their own sources of capital, this can be a high-leverage opportunity window to work with, rather than work through, the government. And once the forest fire settles down, there will be sunlight for new ideas to grow rapidly.

Here are three kernels of solutions to the problems:

1. Create a government fund that moves at startup speed

The institutions vs individuals dynamic will play out in science. This is good because we need more startup-like energy and less organizational drag. Funding will always be the crux. We need mechanisms that combine venture capital's agility with government-scale convening and purchasing power. We could create a Fund for Accelerating Applied Research, matched by philanthropy and private money, which leverages existing fieldbuilding organizations to discover, cultivate and support teams of practitioners. For more information, see: Academia 2.0 Can look more like startups.

2. Empower the Progress Studies movement

There is a vibrant, organic community that studies the people and forces behind great leaps of progress. Nobody anointed them or certified them, they are independent writers and young organizations that just got to work. At first I ignored them, but I’ve come to love them because engineering growth in target subfields is inherently a Progress Studies challenge. We need to formally recognize and fund this emerging field of Progress Studies, creating an official platform for that includes voices like Jason Crawford (Roots of Progress), Nadia Asparouhova, Alec Stapp (IFP), Eric Gilliam , Kanjun+Michael (Science++) and other leading thinkers.

3. Launch experiments toward elite universities 2.0

Instead of fiddling with who gets into Harvard, we need to build new and better Harvards. America's failure to create elite universities in the past 70 years is an embarrassment. We must embrace the uncertainty of education in the post-AI era by encouraging broad but substantial experimentation. While President-elect Trump's proposed free online American Academy sounds awesome, we also need new models for intensive, hands-on training of elite talent.

Conclusion: Adapting at Technology's Pace

In the AI era, any organization that can't keep pace with technological change is doomed. Large institutions are naturally disadvantaged in this race condition, but when an organization can be both big and nimble, then you get a generational performance like Amazon’s pivot from books to compute.

While America's scientific enterprise remains anchored to universities and federal labs, we have an opportunity to build something better. Through federal reform and ecosystem development, we can create problem-focused, founder-led research organizations that outcompete the status quo on cost, accountability, and speed to discovery.

The forest fire is coming. But like the Frosted Elfin butterfly or the seeds of a lodgepole pine tree, some of our most beautiful innovations might be ready to emerge.

This weekend before the election, my family and I drove through rural New Hampshire. As a swing state, the roads are covered with masses of political signs, most of which are civil or at least funny. What’s it like to live next to someone who is voting for the Other Person? Even though the rhetoric du jour is “save democracy!” I worry that we’re forgetting to be friendly to somebody with the Other’s road signs. This is a horrific thought of course, because truly saving democracy is connecting as humans with people you may not agree with, casting your vote, then getting back to work together the day after the election. In rural New Hampshire, there are practical reasons to maintain respectful and friendly relationships, but those practicalities are increasingly absent in digital and city life. So, in the hope of role-modeling the respectful discourse needed to humanize the Other and save democracy, I’ll trust you with my politics. Then let’s finish this election and continue building things together. -Dan

Sometime in 2012, I’m walking out of a bar to get in a fight. This is funny because I don’t even like bars and I like the idea of getting hurt at one even less. But at this moment I’m with a troublemaker friend, charismatic and handsome in a big night out, who likely made the wrong comment to the wrong guy’s girlfriend. Or it could have been the weird phenomenon of stocky guys picking fights with tall guys. I don’t know what the exact stupidity was, but I’m not bailing on a friend. So the handful of men involved with this primal display fan out on the street, and the two main bucks square off. Everybody is waiting for the first punch to be thrown.

On the beat for the first drunken haymaker, an Uber pulls up and my friend Eric pops out. “Sorry I’m late everybody!” Everybody freezes. Magnificently oblivious, Eric scampers up to the two men in the middle. He hugs our friend then jovially extends his hand to the other guy. “Well, Hi! I’m Eric! What’s your name?”

The other combatant is bewildered and disarmed. “Um, Hi, I’m Brock.” Then shakes Eric’s hand.

And that was it. Then, we all introduce ourselves … and just hang out. After a cigarette or two worth of time, we all go our separate ways.

I think a lot about Eric’s diffusion of that tension. It’s as beautiful as it is comical. In the flash of an extended hand and an earnest smile, the mysterious enemy is instantly just another drunk dude with a name and a bad story of bar shenanigans. We got into that situation because of snowballing emotional momentum that nobody really wanted. Even if Eric had properly read the body language and intentionally done something to break up the fight (he confirms this was purely a naive accident), he couldn’t have done a better mediation if he tried.  

Well, it’s election season here in America, and unfortunately Eric can’t pop out of the car for this one. But it doesn’t need to be that hard to extend a hand and talk to each other.

I wonder how many hardcore Kamala voters know a single Trump voter, and visa versa. We all know there are radicalization forces that push us further away: social apps, the SAASification of political machines (how many spam texts do you get? I get 5-10 a day from Kamala thanks to donations in 2016) and media companies making money from the worst fringes. Yet how many people earnestly discuss their moderate views? I think a concerning majority of people don’t share their views out of fear for negative feedback. 

In finance/tech speak, the term for posting politically is an “asymmetric-negative bet.” Becoming an outcast for a political post is a huge cost for a very minor gain. So most professionals do the rational thing and stay quiet. But now we’re in a Tragedy of the Commons situation because the only people willing to engage with the democratic process are attention-seekers and extremists.

Especially as a person working in climate, I wonder what the blowback would be for someone who calmly said that they took their vote seriously, did extensive research, had long conversations with people different from them and then concluded their vote is best cast for Trump.

Well, let’s run the experiment. This year, after a lifelong all-blue voting streak, I’m voting for Donald Trump. AMA.

More than the Presidential vote, I’m voting for depolarization. I hope this letter holds more weight than my vote. No matter what, I’m here to give everyone my best.

I’m not trying to convince you of anything aside from the importance of civil open-minded communication. It’s the day before the election and I know firsthand that if you feel disgust at Trump, it won’t resolve in 24 hours. Also, it’s warm, personal connections that helped me change my mind, not the internet. I’m just sharing this to do my part for what I think respectful democracy can look like.

No matter who wins, I’ll cheer in January and move on. We’re Americans first and foremost. I have smart and wonderful friends, likely many of my colleagues, who deeply believe in Kamala. So if she wins, I’ll charge forward with them and do my best for an optimistic future. I hope to collaborate with either Administration.

If you are curious about my take, I have a few high-level comments below. 

Would I write this if I was voting for Harris? Yes, if I felt I had something to contribute. There are already a few good writeups out there, and a quick skim on X shows some respectful pitches for a Kamala vote which I appreciate. But I do think it’s important to humanize a viewpoint that is under-represented in science/academia. This is how we build bipartisan efforts that welcome in everybody.

How on Earth can a person who works in climate vote Trump?

To get the quick insults over with, no I’m not a racist, misogynist, Russian asset, AIPAC hack, populist, pundit, low-information voter, low-IQ voter, Trumpet, weirdo, or whatever other mean phrases float around the internet. Please no more name-calling. Take me at my word that I’m a nice guy who works hard to be good to others.

One of my principles is to judge actions, not people. So I have no intrinsic judgements on either presidential candidate. I think they both can be great and I’m only going to react to the campaigns. At the core, I think the Harris/Walz campaign underperformed and Trump/Vance campaign impressed.

The irony is that Harris/Walz ran a campaign based on vibes and protest that feels like Trump2016, while Trump/Vance ran a surprisingly adult campaign. The only real policy I’ve heard from Harris/Walz is Taxes on Tips (lifted from Trump), Tax Unrealized Gains (so flawed even Noah Smith had to roast it) and a page of financial promises to black men in an effort to court their vote. In contrast, and in defiance of stereotypes, the Trump/Vance put out surprising amount of substance that treats their audience as smart. 

The bullet grazed Trump’s ear in July because was looking to the left to point at a graph on a slide about immigration. This reminds me of the subtle but deep optimism Ross Perot must have had in the American political consumer when he bought 30-minute slots on C-SPAN to walk through his slides in 1992. My family and I laughed at Perot at the time, now I’m pretty touched. Similarly, I can’t believe I’m saying this, but Trump was really good on this economics interview: I’m still not sure if he’s technically correct, but he pushes a coherent-sounding thesis and negotiates a debate-heavy interview.

Vance started rocky with going straight to the reactive “childless cat lady” snark, but by the end of the campaign you can see why he was chosen. He is smart and I value his skillset. I listened to him talk about the bond market and international treasuries for 10 minutes on the Tucker Carlson podcast. This impressed me because I worry very much about this topic and I don’t see the Harris/Walz camp addressing it. When a potential adversaries holds the power to crash our currency, and the looming fiscal impossibility of boomer retirement liabilities makes the USD an increasingly questionable asset, we need a person, no matter how impolite, who is qualified to lead experts. 

(yes, ugh, this is a clickbaity title… the link should go to the point of economic discussion.)

The Kamala/Walz campaign turned me off by courting the intellectual voter but refusing to engage dynamically and on any substantial issue. I make space for the argument that she’s playing the campaign extremely cautiously, and might be stronger as President, but my choice as a voter is to emphasize the campaign. I saw statistics in a paywalled PDF that showed a paradox that the majority of college-educated people prefer Trump’s policies but will vote Harris anyway. So we got Brat Summer and “Weird” and “Harris. Walz. Obviously.” swag, and a disappointing avoidance of anything but the friendliest interviews.  Aside from abortion rights, going from just the Harris/Walz campaign, I just don’t know what I would be voting for.

The Harris/Walz opacity on policy mirrors what my conservative friends gripe about with climate topics. My experience is that conservatives don’t doubt climate problems, they doubt the current solutions. And they’re all sick of being treated like they are dumb. The analogy here is Jon Stewart versus John Oliver: Stewart was a great talking head for progressives because his intelligent exasperation flavor of comedy has a broad appeal, whereas Oliver’s tendency to condescension is polarizing.

Lastly, and most importantly, I hear my Kamala-voting friends and family on the abortion point. It was honestly my last sticking point too and I get why women bias towards Harris. I identify as strongly pro-choice and I feel their nervousness, but I also think “reproductive freedom” is the only strongest leverage point Harris/Walz have so they are digging in hard with all tricks. To the best of my research, Trump has distanced himself from the extreme stances on abortion. So my best guess is that the most likely steady-state outcome is the status quo we have today, effectively leaving it to the states, and it will remain regardless of the who wins the presidency.

So to wrap up, here are the issues I see:

• Cyclicality: Cancellation fears by moderates have enabled excesses of the Left, and it’s time for a cyclical shift. San Francisco, Chicago, New York and Portland are cities suffering from these excesses and would benefit from breaking the political monoculture. The past 20 years have brought magnificent progress for LQBT and DEI efforts, and I cheer for these. My vote is based on the belief that a peaceful cyclical shift will lock in this position of the Overton Window, while pushing too much further Left will risk the gains made.

• Financial: Government spending has to be brought under control if we are to balance the budget. It not, we risk becoming Argentina, printing money to nowhere while other countries disregard the USD. Consider that Social Security was invented when the ratio of workers to retirees was 30-to-1, now it’s a 3-1 workers-boomers ratio. If there is not a narrative of US stability that investors believe, the results could be catastrophic.

• Maslow is real: Climate policy means nothing if we have a failed currency and unsafe place to do the work.

• International relations: Yes it makes people uncomfortable that Trump and Putin have a direct personal line, but Biden calling Xi a “Thug” for four years didn’t accomplish anything. I’ll vote for human connection.

• The crux of “The Dress” olive/blue: To some, Trump is a fascist monster. To others, he is the only path to global stability. I used to believe the former, but today I’m not convinced by either of these takes, so they zero out for me. Similarly, conservative communities have their own choice of how to view Kamala. I choose to see either option as just another US President.

• Depolarizing the media: Once you step off the Kamala road, you get pretty concerned by the media coordination and see why public trust in media is plummeting. This was Trump’s 2016 favorite refrain, but I think the reaction to him has ironically increased the coordination to a point that most people can see. The sudden effort to define masculinity was a pretty clear canary moment for me.

If I’m wrong, I’m wrong. This is where my research and conversations got me in this election cycle. And if we disagree, we can disagree constructively. This is how science and progress happens.

At the end of the decade, we need progress more than vibes. So I leave you with this remarkable moment of a skyscraper-sized rocket landing back on Earth with a precision measurable in centimeters. Say whatever you want about the guy personally, but this moment not have happened without Elon Musk. Let’s wrap up this election and move forward together.

Author’s note: This is one of those essays that starts simple and ends up big. If you make it all the way through, here are the three ideas and the conclusion: (1) Private groups like domain-specific venture capitalists and non-profit scientific organizations, which I collectively call fieldbuilders, are the best model we have for spurring innovation ecosystems filled with empowered, collaborative teams. (2) There is a widening hole in the status quo. Universities are weakening rapidly and government models for advanced research were developed in the 50s in a radically different economic public/private balance. So it’s essential to prototype scalable models for future talent development and research. (3) The unit of team research, if not done in a university or a startup, requires a repeatable structure to accept and deploy capital. Together, I argue that the government could use a fund-of-funds approach to fund the fieldbuilders to fund —and cultivate— the teams doing cutting-edge research.

Enjoy and I look forward to discussions. -Dan

How do you price something that has little value today, but immense value in the future? -Cooper Rinzler

During the 1930s, while most leaders were hunkering down for the Great Depression, William McKnight went on a hiring spree. McKnight was the President of Minnesota Mining and Manufacturing (3M), and in this era 3M ventured beyond its traditional work in sandpaper and abrasives. They achieved notable blockbusters like Scotchlite Reflective Sheeting (still used on traffic signs today) and Scotch Tape in its iconic “snail” dispenser (still on my desk today). When the economy did improve, 3M was filled with talent to capitalize on the upswing. For this exceptional leadership, McKnight is often referenced in business education as a gold-standard for taking a functionally optimistic longview in the face of a macro-scale downturn. 

If McKnight were alive today, he would say we must invest through the market dip in biotechnology. My unofficial guess is that Ginkgo lets go of 25-50% and once-shiny startups like Cemvita who fail to raise their Series B1 need to layoff the majority of their teams. Amyris went bankrupt last year, most (all?) SynBio SPACs are down 80-99% and the Fierce Biotech Layoff Tracker has had a busy 2024. Outside of pharma and agriculture, biotechnology struggles to match its immense potential of a platform for atomically precise and infinitely scalable technologies to short-term funding options.

So I propose we make a USA Applied Research Acceleration Fund (ARAF). This is not a new institution but a monetary vehicle akin to a fund-of-funds available to citizens who are already doing the relevant work of cultivating teams and scientific progress. It’s to combine the U.S. government’s excellence at convening resources with the civilian ecosystem’s advantages of speed and autonomy. While anybody could apply to deploy funds from an ARAF, I imagine Venture Capitalists (VCs) and domain-specific research non-profits (aka, fieldbuilders like Homeworld Collective) would be most competitive and aligned.  In practice, an ARAF would look something like:

1. Pick the problems: Use the convening power of the Government to agree on a few frontier-level challenges (eg: gigaton-scale carbon capture, brownfield metal mining, quantum engineering) that are bottlenecked by science risk. Critically, use thought leadership to cut through a vague theme into specific problems that need to be solved.

2. Institutionalize the new growth curve: Create a new growth curve for problem-focused ventures (profit or non-profit) that charge right into science risk. I will argue the right growth curve is a sigmoid, as opposed to the standard exponential curve that all startups implicitly utilize.

3. Open the competition for the funds: Funders with domain expertise will compete to manage this capital. These funds will accept the standard 2% management fee but shall only be used for underwriting science risk. 

4. Let professionals deploy the grants: The funds will diligence the team and the potential impact of the technology if it works. Taxpayers and ARAF Leadership can monitor performance using the standard venture capital LP accountability structure with memo writing, 

“ARAF” is a terrible name. ARAF a clear placeholder to remind us that these are ideas in motion. That said, I’m beginning to see what the ideal next iteration of western academia, and its relation to science and industry, needs to be.

Four foundations for the ARAF argument in the context of biotechnology

Because I have seen some fields work very well and some fields function poorly, I look at the progress of scientific fields as something that can be engineered (see: “What are the most productive communities?”). There is a four-part groundwork to the eventual solution of an ARAF.

• Major layoffs are blowing a hole in the Synthetic Biology field: The public companies with “Synthetic Biology” branding are down in a big way: LanzaTech stock is down 80%, Ginkgo/Zymergen down 97%, Berkeley Lights down 95%. In the private markets, startups like MicroByre and Scifi Foods shut down. While the best of this vintage of companies are continuing to fundraise and grow, there is now more talent than open roles. In conversation with Matias Kaplan looking at the FierceBiotech Layoff Tracker, we roughly estimate 5,000 jobs have already been cut and I wouldn’t be surprised if there were another 5,000-10,000 layoffs on the horizon. Depending on how you do the math of PhD-level biotechnologists doing things other than drug development or corn optimization, this might be 25-50% of the US SynBio workforce. 

• Biotechnology is a critical pillar of future technology. Maintaining American leadership at the cutting edge of biotechnology is of national importance. With the surplus of bioengineering talent, we don’t have an elastic structure that can continue to deploy and develop biotech talent. Think of the loss if McKnight’s 3M had let their chemical engineers turn into subsistence farmers. There may also be a negative present-day case study: The failure to invest in long-term workforce is playing out in the nuclear sector as it scrambles for a shortage of trained staff.

• Government funding is critical for innovation yet cannot scale wide nor small nor fast: Less than 17% of the CHIPS/IRA was deployed, and will bias towards single large lump sums (Politico, May 2024). There are many fantastic, mission-driven people working inside the government to make it support innovation better, but there are simply not enough empowered individuals with discretionary budgets. This leads to science-by-committee, which leads to a bias to bulky grants rather than nimble funding (see Derek Thompson’s America Needs a New Scientific Revolution). To give a sense of current scale, the Department of Energy makes ~2000 contracts a year for $42B and ~900 grants a year for a total of $2B with roughly 2,300 staff. Universities have been the primary recipient of this funding, but I think that this is changing. The rapid growth and experimentation of independent research institutes show increasing capability to fund the problem-focused teams directly, without a big intermediate like a university, which is both faster and more efficient.

• Every good deeptech VC has its own hack for working with non-dilutive funds. While most investors ride (mooch) on trends, some VCs actually create movements (eg: Founder-Led Bio). When done correctly, great investors create a virtuous cycle of talent creation reinforcing value creation. This work creates natural collaboration points between VCs and grantmaking organizations, but everyone does it a different way. Back in my design days at IDEO, “look for hacks” was one of our ethnographic research guidelines: the parallel evolution of many investors making up their own strategies of investing in companies alongside grants looks like hacks to me. 

What we need is a mechanism that enables biotechnology to continue growing despite fluctuations of short-term markets. We want a scalable funding and management system that grows the talent along with the technology, led by credible operators with relevant expertise, with constant encouragement for small teams to tackle big problems. There must be accountability in the short term but supportiveness to progress in the long term. Scale comes from a repeatable structure that incentivizes talent and holds everyone accountable to good behavior. Even when functioning perfectly, universities satisfied some but not all of the desiderata, but academia is progressively becoming a harder place to do good research.

Universities are sick and getting sicker. The undergraduate numbers tell a powerful story: US colleges are closing at a rate of one a week (going from 4,599 degree-granting in 2010 to 3,932 in 2021), meaning there is a growing hole for training the next generation of talent. Dropping enrollment is also indicative of university leadership failing to create value for the students: for example, I can’t see how $92,167/yr at Tufts University is worth it. So while I focus on pricing and prosecuting biotechnology research, I feel the urgency to find our next winning recipe to train a talented and empowered workforce.

The fieldbuilding dollars for SynBio/Biotech can’t just go to universities and Ginkgo anymore. 

Step One: Agree on Problems that must be solved

Let’s say we create the ARAF, a billion-dollar pile of cash to be deployed into the US innovation ecosystem. Where does it go?

We know that capital must be focused in order to create progress. Without structure, entropy takes over and while the ecosystem might heat up for a bit, little forward momentum is generated. Capital deployed on themes alone is likely to have little effect – aka “Climate Emergency!” or “Web3.0” – whereas thesis-driven investment can outperform if the team is good enough. Creating a progress movement centered around major problems is a great way to structure the capital deployment.

We must create a problem-centric culture at the national scale. Over the past year, Homeworld has been quietly developing its Problem Statement Repository, and in our first blog announcement we reference the leadership of Heilmeier (DARPA) and Hamming (Bell Labs) who created much of our modern language about tackling problems. Homeworld’s thesis is that a broader discussion of problems is the first step to creating substantial solutions, and our progress at Homeworld confirms my hunch that problems are the correct battleflag for an ecosystem to rally around.

The fields that have been growing most –AI and genetic engineering– rarely talk explicitly about problems because good problems are so implicit in how those fields operate. If you see the Mistral Large 2 announcement, two days after the Meta Llama 3.1 announcement, you see that they are implicitly talking about solving the same problems, measured by the same benchmarks. For fields like biology, which is less linked to benchmarks, it takes a bit of extra work to explicitly state key problems.

So the first step is to use our existing expertise network to find the themes (climate biotech, quantum engineering, artificial intelligence, green chemistry ec), then unpack them into problems. The ontology might look something like:

Theme: Neuroscience

• Problem: How to achieve whole-brain connectomics at synaptic resolution in mammalian brains

• Problem: How to scalably access neuronal dynamics in living brain

  • Subproblem: Develop minimally invasive nanoelectrode arrays for high-density neural recording

• Problem: How to discover universal principles of crossing the blood-brain barrier?

Theme: Climate Biotech

• Problem: How to engineer nitrogen-fixing symbioses in non-leguminous crops?

• Problem: How to develop bioengineered consortia for efficient bioremediation of heavy metal contamination in abandoned mines?

• Problem: How to scale up cell-free biomanufacturing for sustainable production of high-value compounds?

(Note: These are purely meant to be illustrative)

After we build agreement on big problems in a few themes, teams will start proposing solutions. Now we need a way to fund those different teams. Capital needs a “growth curve” as some sort of model of expectations across time. Surely we can do better than the status quo of academia and startups.

Step Two: Create a New Growth Curve

Right now, every visionary scientist is forced into either academia or startups. These are the only repeatable methods we have¹. This is a duopoly and the harm is that many great talents are robbed of agency or get locked into malformed startups. The benefits of creating more paths is that a whole latent class of talent, the scientist-founder, can be unlocked.
Developing the Problem-Centric Founder

Every startup has to immediately tell an exponential growth story. If there isn’t a exponential story, then that science has to be done in a university or subsist off government grants and contracts. This is a status quo with significant room for improvement:  If we want funding allocators to invest in teams that tackle science risk, as opposed to market risk or technology risk, it means creating and disseminating an alternative growth curve that funding allocators can understand.

The practice of selectively funding teams with an exponential growth model is only fifty years old but has deployed over a trillion dollars in the US in the past five years alone (statista). This is the venture capital industry and it operateswith just 5,000-10,000 people (link). Exponential growth – and its corresponding dreams of billions– is why a few venture capital firms have yielded staggering results. It is also why the vast majority fail to breakeven. But as I’d explored in the Problem-Centric Founder essay, the exponential assumption just doesn’t make sense when the entire startup depends on science that is fundamentally unpredictable.

What we need is a precursor to the exponential growth curve. We need something that captures a step function of pre- and post- scientific risk. The sigmoid, famous for its use in machine learning to map a system’s output into a binary “go” or “no-go” value, is appropriate here. In Developing the Problem-Centric Founder, I shared our experience as non-profit founders at Homeworld. We are a fieldbuildign organization, not necessarily embarking on one single scientific endeavor, but I believe our experience generalizes well.

The intuition behind the sigmoid curve is that the primary science risk must be tackled before making assumptions on the future trajectory. In this model, there are three possible outcomes: 

1. If the science doesn’t work on your given timeline, just open source what you have and move on with your life. Ditto if the result is good but not great. The faster, the better! Let’s count this as a true negative and make sure everybody grows from it.

2. If the science creates an important result but still can’t fit on the exponential curve, then find a non-profit framing that would support the problem at some informed baseline of needed funds. Re-raising ARAF funds on another sigmoid with a new science-risk target is a fine outcome. 

3. If the science works out and has potential to work as a startup, raise investment and charge forward. 

Let me emphasize the celebration of true negatives, as Bell Labs effectively did this model a century ago. In The Idea Factory, you’ll read that the organization’s spirit was to consider true negatives a form of success: The people in a valiant but failed project would recycle into other teams while they ponder on their next project. Today we have a beautiful ecosystem of startups and startup supporters that give credit to founders for trying a company that fails, but the knowledge tends to be lost. The ideal is to recreate this Bell Labs energy that recycles and shares the WHY behind the true negative outcome, but across the larger science community.

So how do you craft a sigmoid growth function for a science-centric team? Easy: The exact same way investors today work with teams to craft their exponential growth functions.

Step Three: Leverage gold-standard existing investment practices

Venture Capitalists (VCs) and non-profit Fieldbuilders are well positioned to be the groups that the ARAF empowers to find and support the teams tackling science risk.

I am not here to fawn on VCs. In general, they do that for themselves just fine. But we have to appreciate that the best venture capitalists have been catalytic to technology and talent development. 

Part of a good VC’s job is to help the founder craft meaningful milestones to a startup’s growth. This is a true value-add service, especially when the investor is actively engaged with helping navigate the startup through actually accomplishing those milestones. Crafting a good inflection point on a sigmoid is, in principle, just like any milestone: it requires executive and technical expertise to frame an achievable and valuable goal. When evaluating whether to invest or not, the good venture capitalists will write memos of their investment decision. Science could benefit from the practice of memo writing.

Bessemer Venture Partners does an exemplary job of sharing their portfolio (and comical anti-portfolio) investment memos. Writing these memos is done for accountability to the limited partners of the VC firm to explain the what and the why of the investment decisions (more examples here).

The practice of rigorous memo writing and investment committee meetings feels fully applicable to deploying scientific grants. As with everything, the challenge is in the implementation. 

Step Four: Open the ARAF competition and let anybody apply

We now have the structure of the science (themes/problems/solutions), a funding structure per team (the sigmoid) and the reporting for assessing teams to receive funding (the VC memo). The last step is to fund the funders.

Anybody can apply to be an allocator for the billion-dollar Applied Research Acceleration Fund. Here are some possible terms.

• Fees: The ARAF will pay the standard VC management fees of 2%. 

• Limitations on investable targets: The ARAF funds under management must be relevant to the Priority Problems. Also, funds cannot go to existing for-profit companies. 

• Reporting: An ARAF deployment of capital requires an investor memo explaining the addressed Priority Problem, the science risk undertaken by the team and the inflection point that the ARAF allocators believe can be addressed with the given funding. 

• Non-exclusivity: If the ARAF Allocator is an investor in companies, ARAF funding must precede any venture investment in the team, and the investor deploying ARAF makes no formal claim on exclusivity of funding rights into the team. Put simply, the ARAF funding is an opportunity to build relationships, not lock in talent.

• Allocator Expertise: The ARAF Allocator must demonstrate domain expertise relevant to the list of priority problems by the ARAF fund. Expertise in scaling technology into successful companies is an advantage but not a requirement. 

• Follow-on: A grantee of ARAF funding can receive a follow-on ARAF, but would require a new diligence memo explaining why the new milestone is a fit for the ARAF program. 

• Assessment: ARAF Allocators will be evaluated on the quality of their analyses and performance of grantees. Note that discovering and publishing a true negative is considered an honorable win. An ARAF allocator which has a 100% success rate in turning grantees into companies wouldn’t necessarily look good. 

Venture Capital funds would naturally be advantaged in competing for this, but they would now face viable competition in a post-ARAF world as the fieldbuilding orgs develop. We are in a high-growth era for problem-focused research non-profits. Consider a few examples since 2020: 

• Fast Grants deployed $50M for covid research.

• Impetus Grants deployed $26M for longevity research.

• Unitary Fund has given away more than 100 microgrants for quantum technology.

• Spark Climate has done multlipe rounds of grants for mitigating methane.

• Homeworld Collective has deployed $1.3M for climate biotech research.

All of these organizations leverage their reputation and expertise to be good allocators of capital within their domains. They were also done with shockingly small teams of 2-10 problem-focused people empowered by philanthropy. I believe the experiments into fieldbuilding organizations, initially funded by philanthropy, are now ready for government funds.

Conclusion: Major shifts in funding philosophy have happened before and must happen again

An important but forgotten fact of Silicon Valley is that the first LP investment into a VC fund was a university. Traditional money managers in New York balked at the idea that a fund would invest in small teams pitching ideas from fancy slideshows. But in 1978, the Stanford Endowment invested in Sequoia Capital. It took the perspective of funding allocators sitting inside a university to see the synergy with a firm that might one day take an interest in Stanford graduates. Besides being an excellent financial result for the Stanford endowment (eg, investments to Apple 1978 and to Cisco in 1987), this had a fantastic cultural impact on the Stanford community: 50 years later, Stanford is the best university in the world for startups.

Analogous to Stanford’s incentives to invest in Sequoia down the street, governments have reason to invest locally, too. In 1978, the same year Stanford Endowment funded Sequoia Capital, the Commonwealth of Massachusetts established Massachusetts Technology Development Corporation (MTDC), now known as MassVentures. This initiative began with a $2 million appropriation from the state legislature, creating an evergreen fund that has been operating for over 45 years. MassVentures has since invested in over 150 companies, creating thousands of jobs and generating significant returns for the state. At the federal level, we're seeing innovative programs like the National Science Foundation's EArly-concept Grants for Exploratory Research (EAGER), which provides up to $300,000 for high-risk, high-reward research, and the NSF Regional Innovation Engines program, which aims to advance critical technologies and address national challenges through regional coalitions. Additionally, in October 2022, the bipartisan CHIPS and Science Act authorized the creation of the Foundation for Energy Security and Innovation (FESI), a private non-profit designed to channel private-sector investments into Department of Energy priorities. These examples demonstrate a growing recognition of the need for more flexible, venture-like funding mechanisms in the public sector to drive innovation.

So an ARAF is becoming more doable by the day. I argue that we need to take this weird idea seriously because the status quo is weaker than we think it is. PhD students are waking up to the fact that they’re being trained for professorship jobs that don’t exist. It would be a pity to lose our progress in biotechnology because our best microbial engineers can only get jobs screening cancer drugs.  We depend on catalytic government funding but we can’t be bottlenecked on procurement limitations and recruitment of Program Managers. 

The economy has changed a lot since the US Government started funding cutting-edge science in the era of Vannevar Bush. When DARPA was created in 1958, the salary of the US Treasury Secretary was $25,000 ($250,000 in 2024 dollars), which was both a higher salary and far more prestigious role than any comparable job on Wall Street. Today, the US Treasury Secretary Yellen ($221,400 in 2023) earns 1/100th the salary of David Solomon ($25M in 2023) as the CEO of Goldman Sachs. This is of course an extreme example of a government-to-industry comparison, but I bet a DARPA salary was upper echelon in the 60s and 70s. Government deployment of capital does not scale well if it’s gated on recruiting talent that is no longer maximally incentivized.

So I think the solution is not to create bigger institutions or to pack more into existing universities. Instead, I think it’s better to pick the important goalposts and empower the community of innovators to self-organize to take their best shots. These teams are already out there today, supported by philanthropy to test their models for encouraging innovation, and now I think we can say these models work. 

So if the Western world says we value freedom, autonomy and incentives, we need to look to where we’ve been most successful. Our strengths today are not in creating large bureaucracies, our strengths are creating environments for which small teams are incentivized to grow into big teams. A “fund-of-funds” funding philosophy from the US government to fund fieldbuilders/VCs to then fund small teams to do cutting-edge research is just crazy enough to work.

Acknowledgments/Inspiration

• Paul Reginato and the Homeworld Team

• Niko McCarty for constantly helping me grow as a writer and detailed feedback on this piece.

• Nazish Jeffery for sending great feedback and questions

• Mike Fisher’s BAPO piece on Day One, specifically his BAPO Ventures concept.

• Ben Reinhardt and SpecTech

• David Lang and Experiment

• Isha Data and New Harvest

• Martin Borch Jensen and Norn Group

• Jason Crawford and Roots of Progress

• Braden Tierney and the Two Frontiers Project

• Dom Falcao and Deep Science Ventures 

• Derek Thompson’s writing at the Atlantic

• Michael Nielson and Kanjun Qiu’s monster metascience piece

• Adam Mastroianni’s essay on The Rise and Fall of Peer Review, which questions a sacred foundation of modern science. It is a great reminder that our status quo is far from optimal. 

• Seeing like a State,  by a recommendation in one of Tom Kalil’s interviews. The first half does a great job telling the stories of top-down failures of management. Specifically, the juxtaposition of Jane Jacobs’ study of cities by walking the alleyways versus Le Corbusier’s city architecture strategy of flying over the space is an apt metaphor. 

• Juan Benet’s talk on Decentralized Investment Structures in 2022. I think this is a great talk arguing for the value of expert intermediaries when allocating funding in complex spaces. My takeaway is that there is an upper bound of how many grants a single team can make, so we can learn from the hierarchical funding structures of the private markets to fund science better. 

Imagine if Michael Phelps decided today to create a competitor to the Olympics. With just four weeks before the Paris 2024 Opening Ceremony, Phelps rents the largest event space still available in Paris and calls it the Michael Phelps Invitational. Wearing his 23 gold medals around his neck, he goes to NBC and pours out five million dollars of cash onto the announcers’ table to say “it’s criminal that my fellow Olympians work so hard for zero pay and the hollow promise of glory. I will pay $50k to every Olympian who quits their event to instead walk across the street and compete at my event. At the Michael Phelps Invitational, winners get a million dollars.” What would you feel?

My first reaction is to protect the sacredness of the Olympics. I love the Olympics and everything it stands for. It’s not about they money, I’d say, it’s about the honor of representing your country on the biggest stage in the world! It’s the Olympics because … well, it’s the OLYMPICS!

This is more than a thought experiment: This drama is happening right now in the world of professional jiu jitsu. One of the most famous grappling athletes in the world, Craig Jones, used his star power to set up the new Craig Jones Invitational (CJI) right across the street from the most prestigious event in Jiu Jitsu, the Abu Dhabi Combat Club (ADCC) on the same weekend.

On the surface, this story is hilarious. Craig is a unique blend of technical brilliance and comical irreverence. He says “time for nose beers” in post-fight interviews, posts interviews with prostitutes on a hotel bed1, rips off the Chicago Bulls logo for his teams’ (“The B-Team”) competitive matches and his repeated joke that “Jiu jitsu is powered by autism and steroids” led to a line of sold-out waterbottles. He’s like a half-sober Luke Skywalker taking on the Death Star: ADCC fills arenas with tightly polished shows, enjoys brand hegemony, is led by a strong public figure who is respected by the community. Huge names in the sport are publicly thinking through their personal dilemmas with noticeable holes being blown into the ADCC roster. And what fuels this tense narrative is that Jones was once a golden boy of the ADCC arena.

Deeper down, CJI vs ADCC is a fascinating drama about the difficulty of staying on top. It illustrates why dominance based on prestige can collapse if taken for granted due to the magic of positive feedback cycles2. If I were a leader of a Top Ten university watching the biggest brand in a sport shaken suddenly into a defensive scramble, I’d be wringing my hands: can a defected insider burst the bubble of prestige, and if so, can the displaced leading organization recover?

If you know me and/or you’ve read this blog, you know I love jiu jitsu. Its exponential growth since I’ve been engaged in 2010 is (in my opinion) one of the better trends in western society. Jiu jitsu has changed how I think about education and productive communities (see my writing on the importance of “enforceable honesty”). In its purest form, jiu jitsu athletes fit somewhere between entrepreneurs and science professors.

This essay is not about the details of the martial art but about the social dynamics at play when there is a finite amount of top talent and room for one leader. A maximal position, like the title of Most Prestigious Competition in the World, is either stable or unstable. Consider two examples3:

• San Francisco’s position of best tech hub in the world is an unstable equilibrium. The feedback loop is that the best in the world go there because the best in the world are already there. But what if the best in the world go somewhere else? Every time the city leadership does something awful (example 1, example 2, map of closures, etc.), I cringe at the fear of seeing the doom spiral start. Chicago’s $150M incentive package to repopulate downtown (WSJ, NYPost) seem like a desperate stab for a city already in the doom spiral.

• A stable equilibrium would be something like the oligopoly of the Big Four Agriculture Companies. If you want to sell a genetically edited crop as a startup, you have very little choice aside from partnering with them. The practical advantages of scale (eg, government relations and the cash reserves to navigate ~$1B deregulation processes) means that any upstart threat to one of the giants threatens all of the oligopoly, and that any misstep of one of the oligopoly will be filled in by another. Every perturbation to the system seems to return to the same point.

A Gold Medal from ADCC is the biggest accolade in the sport of jiu jitsu. While many productions have sprung up to capitalize on the growth of the sport, most of them either steer clear of ADCC or bow knee to ADCC’s primacy. One such stage is the professional series of WNO professional league, put on by FloGrappling (the biggest media company in the space), and WNO directly builds up to the importance of the ADCC stage. The post-fight interviewers will ask things like: “How do you think your win tonight prepares you for the big stage in ADCC?” ADCC’s brand and history felt untouchable until only weeks ago.

Craig Jones is one of the few with the power, clout and gumption to break the virtuous cycle of ADCC’s prestige. Jones has been beating a drum for years now that it’s a crime that ADCC only pays $10k if you win (video). That means you fight the hardest fights of your life, amplified with the energy of tens of thousands screaming fans, for a paycheck that would barely cover medical bills. One of the Hall of Famers, Xande Ribeiro, said that he’ll stick with ADCC but shared that he invested $2M over his professional career to prepare for ADCC. For Xande, this was likely a profitable investment because of his revenue from selling instructional videos and seminars, but competitors just one rank beneath him are in likely in debt. It illustrates that as the game of jiu jitsu developed around its flagship entity, the flagship entity didn’t developed fast enough in turn. As conditions got harder on the athletes, it made room for Craig Jones to stick his foot in.

Imagine that I open up an institute right across from Harvard. I tell the best science professors that they could quit Harvard and work at the Dan Goodwin Invitational.

• I’ll double your pay overnight and you’ll never be required to teach an intro undergrad course again.

• We’ll drop the 80% overhead costs and save you burning 40% of your career grant writing through a mixture of expert technical staff and AI.

• The leadership will only be other highly credible scientists and people with deployment experience either through companies or scaling successful government projects. The core mission will be creating scientific excellence.

How long would Harvard be Harvard? As soon as it’s no longer the Best Place In The World to be a professor, the new Best Place will suck in the talent. The Harvard strategy of prestige simply doesn’t work if the best don’t want to go to Harvard. Soon, the best students won’t want to go to Harvard and this might already be happening according to the 5% decrease applicants4.

It’s a great season to pile on the trash talk on the Ivy League. Nate Silver publicly giving up on the fancy schools in his essay Just Go To A State School hit me hard. But every recent public statement by Harvard seems to make the Crimson H look worse: the latest fiasco is an essay by the Dean of Social Studies attempting to weigh in on free speech is both appallingly low quality in substance and insultingly chiding in tone. In the short term, leadership for the status quo can rely on the TINA mantra of Wall Street —There Is No Alternative— but I think the ecosystem will be rocked once the first viable alternative arrives.

Top professors are leaving academia: Aviv Regev at Harvard, Hans Clevers at Stanford, Sebastian Seung at Princeton, Daphne Koller at Stanford, and many more, all left for industry where they had higher salaries and higher leverage. Seemay Chou left to create Arcadia. Sam Rodriques left to create FutureHouse. One could argue this is a healthy and normal churn, but my perspective is the conditions of academic professor life are on such a negative trend that it’s unlikely to select for star talent moving forward. Jonathan Wosen at STAT has written great articles on this brain drain, including a snippet from this article.

The [NIH] working group, which launched last year, had been tasked with reenvisioning a system that is under increasing strain. Life science Ph.D. graduates, who have for years complained about low wages and long hours in labs, are skipping postdocs, a temporary period of research and professional development, and going into lucrative industry jobs at historic levels.

Back on the mats, ADCC is on the defensive and has stepped up to match many of Craig Jones’ critiques. Women are paid more, purses for all winners have increased. So the Invisible Hand people who read this could extoll the importance of free market competition making all players better. Maybe this is a scenario in which everybody does win5.

US Universities are on bad trends and are due to have their own “CJI moment” when a viable competitor emerges. That competitor will have a model that is scalable and demonstrably better at one (or both) of universities’ core functions of training and deploying intellectual talent. When that happens, I hope the universities will rise to the challenge to make the necessary leadership/bureaucratic overhauls, as ADCC seems to be doing6. I think it’s inevitable that Universities in ten years will look very different than what they are today, the question is whether it will still be MIT/Harvard/Stanford still leading the way.

And the 2024 ADCC this year will be awesome, I will watch it, but I’ll frankly be watching the CJI feed first.

We can create more options for science founders. 

Right now, every visionary scientist is forced into either academia or startups. These are the only repeatable methods we have1. This is a duopoly and the harm is that many great talents are robbed of agency or get locked into malformed startups. The benefits of creating more paths is that a whole latent class of talent, the scientist-founder, can be unlocked.

I can critique both academia and startups because I’ve done my time in both. I also love them both: Universities are the reason that I’m an American (my parents came here for my dad’s PhD) and co-founding a startup has been the most rewarding parts of my career. But, I’ve also seen the dark sides of both: I’ve seen my postdoc friends go through 200 faculty applications only to arrive in an impossible academic culture of trudging to tenure through awful administrative muck, and I’ve seen wonderful startup founders get permanently warped by getting stuck in a zombie2 company that takes five years to die. 

We can show love to startups and to universities by giving them viable competition. And I think this is a special time to invent a third option.

About today’s duopoly

Startups are the best wealth creation vehicle humanity has invented and the university is the best knowledge creation vehicle humanity has designed. But that doesn’t mean they perfectly span the innovation space. Startups are best suited to scaling technology, not addressing a science risk. Similarly, universities are designed to train talent, not deploy talent3. We need to stop wedging every idea immediately into one of two tracks: let’s call the status quo the entity-focused mindset. 

What we want, instead, is a third path in the innovation economy for a hotshot researcher to entirely pursue solutions to a big problem. Let’s call this the problem-focused mindset. In this new path, it’s the problem that matters, not the entity that initially receives the money. 

If you’re an ambitious scientist today and you need funding for your big idea4, you have your choice of two entities to house your work. You can go academic, and the great thing about this path is that the money can be “free5,” it just takes a long time and is constrained by specific grant opportunities. On the other hand, you can call yourself a startup and the immediate benefit is it’s always open season for savvy investors: if they’re convinced you can make them money, they’ll fund anything quickly.

In the perfect world, addressing science risk is supported with the terms of the NSF but at the speed of venture capital. 

Yes I’m being idealistic but we are making progress toward this vision. The best repeatable solution we have today is to support a scientist-founder type with a Fellowship6. But fellowships are cyclical and scope-constrained. For example, In my domain of climate biotech, Activate is a best-in-class fellowship. But this playbook generally funds individuals, not teams (although I hear Activate is innovating new offerings for the teams too). And the Fellows are still post-docs at universities because there’s no place else to do that research. Tackling science risk still hasn’t fully learned from Silicon Valley’s model of excellence.

Science needs can replicate what makes startup world so great: a big network of small teams operating with freedom, a spirit of collaboration and efficiently funded risk. Founders of these teams help each other in ways that are unique to founders, such as tactical support, emotional support in the hard times, trust to create joint ventures, recognition in the good times. These network effects are tangible, substantial and have been well-studied for years (eg, Professor AnnaLee Saxenian’s 1995 essay on the community effect in Silicon Valley has durable wisdom). And I’m seeing the early stirrings of a new founder community, but they’re not founders of venture-backed startups nor professors. They are a new breed of problem-focused, rather than entity-focused, founders.

Let’s land these meta thoughts with an illustrative example.

A quick founder narrative

Any customer can have a car painted any color that he wants so long as it is black.
- Henry Ford

Any scientist can have a check for any project they want so long as I look cool for investing in them for one year and I make my money back in four.
-The Average VC

You are a black-belt level biologist who can genetically engineer anything with time and money.

Let’s say you go deep into biomanufacturing and come out the other side thinking it’s insane that we’re pushing yeast and ecoli so far beyond their ideal. As Adam Marblestone puts it, you wouldn’t get a cat7, no matter how genetically modified, to pull a plow, so why do we accept this mistake at the microscopic level?

So you talk with the best people in the world and realize that, hey, with us working as an elite team, we could solve a set of common problems that could increase the speed of accessing *all* new microbial hosts, not just cherry picking from ones that happen to be convenient. It could be a major unlock for biotech. That’s a sweet mission!

You go talk with some venture capitalists. Look at my team, you say, we’re the best in the world at unlocking new microbial organisms for the trillion-dollar bioeconomy. “Ehhh,” the VCs say, “What market are you going after? What is the exit strategy?”

And you get frustrated because you’re speaking different languages with your Patagonia-vested tablemates. You are flapping your arms saying there is a clear bottleneck in the field of practice, look at all these famous names confirming the problem we need to solve. The money people are saying “yeah but this SaaS+climate company only has some marketing risk whereas you have worse than engineering risk, you have SCIENCE risk, on TOP of your market risk.”

But maybe you’re winning authenticity points by your persistence, so the funder leans forward on their All Birds and says, “OK, fine, how long do you think it’ll take to reach your first milestone?” to which you earnestly say “Just five years!”

Meeting(s) adjourned.

But, let me tell you that this outcome of rejection is FAR better than another world in which you’re stapled to a sinking ship.

The nasty ending is in the world in which you DO take the money. You fudge a few numbers (“Did I say five years? Oh no I mean two!”). You and the team sprint all-out to the first market opportunity and drift away from the core mission of unlocking novel microbes for everyone. You Google “how to make a business plan,” find out that vanillin seems to have a really high price and a reasonable volume, so you bail on your dream to take a short detour to profitability.

Your mission-driven team is now pretending to be market-driven mercenaries sprinting for some product nobody cares about. You keep telling yourself that “once we reach profitability, then we will be free to go back to our original mission of unlocking biology for everybody.” You limp over the finish line of the first milestone, raise a bit more cash, then finally get to market to learn that buyers of vanillin won’t pay what you’d hoped. Turns out that vanillin is easily detectable as synthetic by carbon isotope analysis and synthetic vanillin is worth a fraction of plant-based vanillin. You now have zero product, zero mission and have lost control of your company due to the terms of your latest funding round. Welcome to startup hell, you are a zombie company that won’t die until your investors say so.

But, thankfully, in this world, you did not take the VC funding. You did not over-rotate on some prestigious yet underqualified advice. You pushed and you pushed and you pushed and you found funders who understood that you are working to solve a problem first and foremost. Instead of small money from VCs that could have turned you into a zombie, you raise serious money from philanthropy that empowered you to create Cultivarium.

In this better world, you are Henry Lee and Nili Ostrov. And you are having a lot of fun doing your science at a scale far beyond that of a professor and with the freedom to steer directly into the science risk that a startup couldn’t stomach.

How do we get more positive examples like this? How do we get more world-class talent to stay true to what they KNOW is the most important problem that needs solving? How do we build an ecosystem of peers, funders and victories?

We’re in era of rapid evolution for how and where we do science

Cultivarium is a Focused Research Organization (FRO). A FRO is a new type of organization designed to operate on the order of a ~5-year, ~$50M time scale. This scale of project was crafted carefully to address a specific gap articulated through extensive enthnography of big-thinking scientists. Today, Convergent Research is the primary developer of FROs, but in principle, anybody can set up a FRO-like entity. These are, by default, non-profit entities that are solving a scientific problem. As there are now multiple FROs that are attracting top talent and philanthropic partnerships, we can say that FROs are the first substantial entry into the problem-centric playbook.

But I’m seeing a strange glitch that’s forming: I now have conversations all the time that sound like “Can you check out my FRO proposal?” What I’d prefer to hear is “Can we discuss this important problem that I think I can solve?” That is, the launch of FROs substantially shifted the Overton Window of what scientists dream is possible (this is an enormous step forward for scientific culture). But, there is now a over-rotation in trying to make everything sound like a $50M, 5-year project, which is getting people sucked back into the entity-first mindset.

A team of three scientist-founders pitching a $50M FRO is pretty similar to a team of three technology-founders pitching a Series C company before they’ve written a line of code. Only in rare situations is this a serious proposition: discovering and cultivating these teams is what makes Convergent special.

So as the problem-focused economy grows, it will become less discrete and more continuous. If a FRO is a $50M organizational structure, then the natural question is to ask what the $5M organizational structure is, or even what the $500k structure. Can there be proto-FROs to derisk core elements of something that becomes full FRO? This begins to feel like startup funding rounds, and this is a good thing. Funding rounds are the natural mapping of money into progress, there needs to be a continuous growth path for all problem-centric efforts to sit on.

So if we are to take the best from the Silicon Valley entity-centric founder and map it into the problem-centric founder, we would see the following:

• Science-risk can be tackled by using non-dilutive funding to allow the problem-centric founder to steer right into the biggest challenge. This is still driven by milestones but allows operational flexibility until the best entity to support the effort becomes clear.

• Fast funding decisions made in weeks not years, in open scope with value-add funders. I talk trash on VCs because most of them are useless. But the great VCs are undeniably excellent catalysts, mostly because they were once leaders of significant efforts themselves.  

• Network effects of small teams on comparable trajectories evolves into repeated pathways to success. This includes a common, minimal legal structure and funding terms that all teams follow. 

• A broadening funder base that understands the milestones of the science and trust in the early stage funders. High-brand early checks from Y-Combinator, Activate or Breakthrough Energy Fellows are powerful indicators that entice the future funders to an organization.  

Building a growth model for problem-focused founders

“Every non-profit should be constantly trying to put itself out of business.”
-Isha Datar, Executive Director of New Harvest

I never expected to be a non-profit founder. Yet here I am, because my co-founder Paul Reginato, and growing team, and I are working to solve what we think is the most important problem: only a fraction of biotech’s potential is being deployed to climate because of opacity of important challenges and the siloed best efforts. To do our best work, Homeworld needs to be a non-profit just like World Wide Web Consortium (W3C) has to be a non-profit to serve the internet. 

When I was a software startup founder, it was because I set out with the intention of founding a startup and we discovered a problem worthy of that entity. But, in contrast, our road to Homeworld Collective was through complete commitment to a problem, the entity came later. By being focused on the mission of growing the field of climate biotech, the non-profit formulation was built around us.

It’s not easy to be a problem-focused founder building a non-profit and I’m very blunt that it took us much longer to raise funds than I expected. Even though we had a clear problem to solve and the authenticity to be the team to solve it, it still took us a long time to raise support for Homeworld Collective. Soon, I hope problem-focused playbook will become more codified, but for now, I think I can do some service by sharing my mental models with the community.

The linear order:

Here is my developing mental model of the arc of funding a problem-centric founder (see image above). If you go the non-profit route, here are the steps I think you’ll go through:

1. Cultural De-Risking. You need to get on the map of funders as a trustworthy and compelling leader. Reputation and legal compliance are of utmost importance to philanthropies, and they have a high noise floor of spammy inbounds.

2. Novelty Moment. You’re doing something radical and beautiful and you’re telling a story that it can’t be done anywhere else but in a non-profit. This is a great time to raise your equivalent of a seed check. 

3. Proof of Concept. You have shown the world that you can actually do something substantial in accordance with your mission and theory of change. You now need to fundraise in preparation for the upcoming Valley of Death.

4. Sophomore Effort. You show you can do it twice, bigger and better. This is important evidence that you are scaling your solution in accordance with your theory of change, and that you are credible to raise larger funds.

5. Valley of Death. You are no longer the shiny new thing, you need to be the performing-exceptionally-well thing. 

6. Three endpoints:

   1. Science is done, published (and all other necessary compliances), and now a startup is the entity to continue solving the problem.

   2. The Valley of Death was successfully crossed and on the other side is reputable non-profit with a strong reason to continue existing.

   3. The Valley of Death was not crossed and the non-profit is sundowned.

Today’s conclusion: the TBD-Corp

If you're going to try, go all the way. There is no other feeling like that. You will be alone with the gods, and the nights will flame with fire. You will ride life straight to perfect laughter. It's the only good fight there is.
-Bukowski

I’m not against startups, I’m against bad startups. I’m not against academia, I’m against academia when it becomes academentia8. 

In the future, I think all efforts facing a science challenge would be better off de-risked through standardized, scientist-friendly non-dilutive terms. We have B-Corps, C-Corps and S-Corps, why not have a TBD-Corp? You raise money in a repeatable form (eg, through a common fiscal sponsor) to get the first bit of progress to solve your problem. If the science looks like it’ll work but will take awhile, you use the evidence you just created to make a research non-profit. Or, if the science worked quickly and you can see a path to market, go straight to build a startup. And if neither the team nor the science works, open source your results and take the W of discovering a true negative.

If done right, this protects and leverages the most important resource: the time capital of our best scientific talent.

A handful of startup funds already do a subtle combo of philanthropy and venture capital this in their own bespoke ways: but it’s generally not publicized. We are still waiting for the “Y-Combinator moment for science,” which will look like standard terms, standard growth playbooks and funders lining out the door to hear the demo days. When this moment comes, the wilting monocrop of the academic status quo will be overrun by a vibrant ecosystem which grows taller and mightier than its predecessor.

See you at the TBD-Corp incubator!

Resources

• Homeworld sits inside a growing pool of peer organizations: Spark, Spectech, New Harvest, Good Food Institute, TripleHelix, Norn Group (creators of Impetus Grants), Experiment.com, Two Frontiers Project, Future House, Align to Innovate, TechMatters, Roots of Progress, Convergent Research, Cultivarium, E11, [c]Worthy, BitsInBio, OpenBioML, and so many more.

• 1517 Flux Capacitor’s approach to funding early projects with $100k on favorable startup-like terms (announcement), which might be the best instantiation of the “TBD-Corp” that I’ve seen.

• Strong recommendation for Erika DeBenedictis’ blog, specifically Quick Start Guide to Research Non-Profits.

• Sam Rodriques from FutureHouse also just published this commentary in Cell.

• Seemay Chou and Prachee Avasthi are doing awesome work Arcadia Science, for example, their publication model of their team science is radically different.

• SpecTech’s BRAINS accelerator is a fantastic first example of cultivating this talent. 

• “A Vision of Metascience” by Michael Nielsen and Kanjun Qiu. A monster read filled with deep thinking.

• Novo Nordisk created the BioInnovation Institute (BII) which offers founder-friendly investment terms that are, as I roughly understand it, a grant that converts into a loan if the project takes form.

• I generally advise everyone to read Ben Reinhardt from SpecTech, Jason Crawford (great recent piece on evolvable science institutions) from Roots of Progress, and follow Sam Arbesman’s Overedge Catalog and read Nadia Asparouhova’s writings.

• Derek Thompson’s 2021 piece America Needs a New Scientific Revolution remains one of my top references.

• Jonathan Wosen has been doing some excellent writing on top researchers leaving university, eg Life scientists’ flight to biotech labs stalls important academic research.

• David Lang does some of my favorite applied work to empowering scientists to ScienceBetter. He has a great recent piece on the unreasonable effectiveness of small grants and one of his great essays is the Hollywood Analogy

Acknowledgements: 

Thank you for helpful conversations and feedback. In random order:

Kenza Samlali, Seemay Chou, Erika DeBenedictis, Pritha Ghosh, Ben Reinhardt, Paul Reginato, Ariana Caiati, Paul Himmelstein, Henry Lee, Niko McCarty, Jim Fruchterman, Jasnam Sidhu, Erin Smith, Devika Thapar, Mark Hansen, Dan Voicu

In this first blog post about Garden Grants, we learn from history to develop the ideas behind this new program. We explore the outsized impact of Y Combinator on startups and Fast Grants on Covid19 research, then announce our collaboration with Experiment.com.
We invite you to started with your journey at the Homeworld Garden Grants website: our first call is for protein engineering and the deadline is October 20, 2023. We will host a public webinar and Q&A on September 28, RSVP here.

Note: this is a mirror of Homeworld Collective’s blog post (link).

Rapid progress in a field needs the right catalyst

In the fall of 2009, I was an electrical engineering PhD student at Stanford who visited Y Combinator for a small gathering. I had barely heard of the place, but there I clearly remember co-founder Paul Graham talking about their recipe for rapidly funding early stage teams with very little evidence. An audience member asked:

"What makes you believe that a team is really serious about working on an idea?" Graham responded, "Easy. They tell everybody what they are working on!"

14 years later, Y Combinator is synonymous with elite silicon valley startups. Even with the impressive portfolio of 4,000 companies (including household names like Airbnb, Stripe, and Dropbox), the legacy of “YC” may be the durable cultural change catalyzed by a single funding mechanism.  YC got the right combination of proposal standardization and engagement incentives that allowed creativity in the software ecosystem to flourish. By 2010, motivated by the outsized success of YC companies, many computer science students at Stanford knew how to pitch a startup. A new culture formed. Mixers and hackathons sprung up, all using the winning language from YC Demo Days. Startups raised money using free templates from YC's website. Even though I never had any further direct contact with YC after the one visit in 2009, my decision to leave my (first attempt at a) PhD to build startups was surely influenced by YC’s cultural effect.

Y Combinator is just one example of funding that changes the culture that changes a field. In science, Fast Grants stunned biomedical funding and Experiment.com is leading the charge on crowdfunding. How might a science funding targeted for the field of climate biotech induce a positive cultural change toward the most impactful efforts?

This is an opportunity for Climate Biotech to move fast and grow things.

Successful climate action requires exponential growth. To reach gigaton carbon capture by 2030, for example, roughly requires a doubling of capacity every year. Biotech has potential to be a medium for difference-of-kind leaps in technology, but the funding needs to encourage big ideas that can be rapidly de-risked.

In climate biotech, everybody speaks the language of the NSF RFP or the ARPA-E FOA because that has been the core of our funding culture. But while those agencies are the bedrock of the climate biotech ecosystem (and are run by excellent people), the existing methods alone cannot create exponential growth in climate biotech. 

Since 2021, our team at Homeworld Collective has met personally with over 500 practitioners and analyzed over 350,000 government-funded projects asking how we can support climate biotech better. Several themes have emerged: 

1. There is a support gap for application-centric efforts that face scientific uncertainty, leaving the best ideas on scientists’ shelves.

2. Many excellent biologists are not working on climate yet because they don't know a good problem to work on or a good collaborator to work with.

3. Even when government funding is available, it can be a 9-month wait to learn whether you’ll receive it. 

4. Fundable areas are mostly determined by the agency, not practitioner community, which may miss truly innovative efforts. 

All of these are solvable.

At Homeworld, we think it's possible for a grantmaking program to fill the funding gap for de-risking early-stage ideas, educate the community about impactful problems, fund projects quickly, and stimulate discourse to prioritize impactful problems. We're launching our first Garden Grants call for translationally-focused protein engineering toward sustainability goals. 

The invention of Fast Grants pushed grantmaking to work better

Fast Grants is the closest thing science has had to a YC-like funding innovation, and it’s no coincidence that funders include Paul Graham and YC alumni John and Patrick Collison.

The Fast Grants team demonstrated the jaw-dropping result that a 2-page proposal and 48 hours of technical review was enough to make $50 million dollars worth of funding decisions for biomedical science. Just like how YC's impact may outlive most of its 4,000 companies, the impact of Fast Grants goes beyond the 260 Covid projects it funded. Fast Grants was a wake-up call to show that scientists are not working on their best ideas because of the frustrating status quo of funding. 

The Fast Grants’ retrospective analysis showed some incredible numbers:

81% of Fast Grants recipients said they’d be more ambitious if they had access to such flexible funding,
78% said they’d change their research topics “a lot” if they had less constraints on their funding, and
62% said they’d work outside their standard field if they could (which the NIH explicitly discourages).

The innovation in Fast Grants was not just about “Fast", it was about building a lean structure to fill gaps. In March 2020, the gap was the NIH couldn't move at the speed of a global pandemic. From our experience, the gaps in climate biotech are funding gaps to de-risk ambitious translational ideas and network gaps between excellent people who would be better if they collaborated. We must invest in community-scale productivity while also supporting specific projects onto successful journeys.

Traditional grantmaking is done in a black box. In most grantmaking calls, the community can’t see the problems being targeted, solutions being proposed, nor what was reviewed favorably. Maintaining confidentiality for intellectual property concerns is part of the explanation for this status quo of opacity, and we believe you can protect confidentiality while promoting more public discourse.

The Garden Grants are meant to be a safe space for ideas to grow

Just as a garden is a safe place for plants to grow and cross-pollinate, we intend for the Garden Grants to be a safe place for ideas to grow and cross-pollinate. The hypothesis of Garden Grants is that separating a grant proposal into a public problem statement and a private solution statement can stimulate community knowledge, discourse and collaboration as part of the grantmaking process. As we socialized this hypothesis, we found an aligned team coming from an unexpected starting point.

Enter Experiment.com, the gold standard crowdfunding platform for science. When we first met the Experiment Team (Denny Luan, Cindy Wu, David Lang), we immediately discovered alignment in our vision for how to fund science better (David's blog is appropriately title Science Better). Experiment has successfully funded over 1,200 projects, so the site has an exquisite amount of subtle optimizations built from helping thousands of scientists -- in topics ranging from interspecies communication to kilometer-deep caving explorations in Alaska-- attract supporters. We, and other granting organizations like Robert Downey Junior’s FootPrint Coalition, see the value in a beautiful public space to communicate project proposals.

Homeworld's vision to separate problem statements from solution statements maps directly to the Experiment.com platform. Experiment centers around a project page -which can be free of enabling disclosure information-- which is the place to convince the world that you are tackling an important, interesting problem. If a reviewer is convinced by the project page, the she'd be motivated to learn about the specific details in a confidential about how you plan to solve that problem. 

First Garden Grants call is for Protein Engineering

Our first call for Garden Grants is for projects that use protein engineering to address a high-impact problem toward a frontier sustainability goal. We chose protein engineering because it’s one of the highest growth areas of biotechnology and proteins are already the catalysts for the biggest carbon fluxes in the world. Proteins could be the centerpoint of a re-imagination of the chemical industry, of sensing or degrading toxic chemicals, or of accelerating carbon capture. 

As Paul Graham had said, when people are excited to tell their friends about what they are working on, they are more likely to succeed. We hope that the process of running the Garden Grants promotes new conversations, new relationships, new skill developments and ultimately transformational new technologies that reinforce each other.

In future posts, we will explain more about problem statements and our focus on protein engineering. In the meantime, you can refer to the Garden Grants Landing Page. Feel free to reach out to grants [at] homeworld [dot] bio if you have any questions or if we can support you in your climate biotech journey.

This is a repost of an essay that Paul Reginato and I wrote on the Homeworld Blog.

We’re thrilled to announce Homeworld Ideas, a $12,500 writing challenge inviting you to share your visions for how biotechnology can enable a sustainable way of life and a thriving biosphere. We welcome all forms of writing, from essays to fiction, from technical experts and non-experts alike. We’re grateful that our friend Niko McCarty (Codon) is leading Homeworld Ideas with us, and that our friends at Pillar VC are contributing runners-up prizes. (Click here for more details, and here to apply.)

You might wonder, why would Homeworld, a field-building non-profit for climate biotech, host a writing challenge? Let’s ask two little fish:

There are these two young fish swimming along and they happen to meet an older fish swimming the other way, who nods at them and says “Morning, boys. How’s the water?” And the two young fish swim on for a bit, and then eventually one of them looks over at the other and goes “What the hell is water?”

Sometimes an idea is so poignant it becomes its own cliche. Most scientists have wondered at the notion that “What I cannot create, I do not understand” (Richard Feynman in the 1980s). In corporate commentary, we still hear that “software is eating the world” (Marc Andreesen, 2009) as the new wave of AI technologies further expands the role of code in our economy. In computing, we wonder about the future of “Moore’s Law” (Gordon Moore, 1998). As we muster the courage to face our biggest collective problems, we remember that “Nothing in life is to be feared, it is only to be understood” (Marie Curie, early 1900s). David Foster Wallace’s 2005 ultraclassic This is Water, from which the story of the fishes is taken, presents an idea so profound and well put that “what is water?” has become a buzzphrase across America reminding us to take stock of our context. Wallace was a magnificent, generational talent who showed us new bounds of how good the art of writing can be. And now, as a climate tech community discoursing on ideas for our collective future, we must realize that the water in which we swim is writing.

It’s writing that stitches evidence into arguments. It’s writing that clarifies our own ideas to ourselves. It’s writing that explains our view of the world. It’s writing that lets us travel between the ideas in each other’s heads. Beyond discourse, writing is an essential skill of self-advocacy. It’s writing that wins us grants. It’s writing that gets us a reply when we reach out. The poignancy and clarity of our writing gets people interested in what’s in our heads, and writing informs people how to help us on our missions. In a very real way, writing is manifesting.

It’s strange, then, that we never truly teach ourselves to swim. In school, we learn to write to prove that we learned what we were told to learn. When we wrote book reports in middle school, we saw writing as a test of comprehension. (See this great lecture to help you break out of this mindset.) In the course of Paul’s technical undergraduate degree, he was never required to write an essay. Even in the best graduate schools for science, we are barely, if at all, taught to write. 

In our journey of growing Homeworld Collective, we have observed the career arcs of hundreds of climate biotech practitioners. It is a harsh truth that some individuals get much farther than others because they are simply better self-advocates. As Dan wrote, one mental model is to think of one’s professional capacity as a blend of craftsmanship and gamesmanship: craftsmanship is how good you are at doing something, and gamesmanship is how good you are at making people care. We used writing to communicate the ideas that got Homeworld Collective funded and started. Writing is an essential component of gamesmanship and, most importantly, writing is a skill that can be taught. 

We’ll say it stronger: Writing MUST be taught and celebrated! If we are to truly encourage a diversity of perspectives and leverage modern connectivity to its fullest potential, then we need to help people, especially in the STEM fields, put down their pipettes long enough to effectively articulate their ideas. Encouraging better writing skills can be a way to help people think better, facilitate a richer discourse, and get better support to practitioners with good ideas.

Here’s the catch: it’s scary to write creatively. You’re probably going to suck at it initially.

It’s OK to be bad. It’s unavoidable.  When you start writing creatively and you cringe at how your ideas hit the page, we urge you to take two minutes to listen to Ira Glass talk about Taste. His point is that there is a gap between your taste and your abilities whenever you start something new, “Because it’s only by actually going through a volume of work that you are actually going to catch up and close that gap.” Homeworld Collective is thrilled to put up a cash prize to encourage you to build the volume of work it will take to become a great writer.

One of our writing heroes has also become a dear friend, and we are thrilled to be co-creating this writing challenge with him. Niko McCarty has synthetic biology degrees from Imperial College and Caltech, and an MA in science journalism. He’s worked his way up from a starving writer in NYC into the writer that every biologist reads. Niko’s mentorship has also made us better writers, and we are inspired by his belief that every person that has great ideas can become a great writer. We’re thrilled that he is leaning into his passion for education to raise others up. 

You can read the details of the essay contest here, but the TL;DR is that we’ve given intentionally minimal guidance on the form your writing must take. It’s okay if your writing is already published elsewhere, you can submit under a pseudonym if you wish, the grand prize is $10,000 with $2,500 of additional prizes thanks to our good friends at Pillar. Have fun and try your hardest to write something you are proud of.

So while Wayne Gretzky skates to where the puck is going and Emily Dickinson dwells in possibility, we encourage you to consider that the pipette may be nothing without the pen. We hope that one day, you too will have produced an idea that shines so beautifully it becomes its own cliche.

Swim on, friends. Reach out if we can help you on your journey.

It's a great time to be a mouse with cancer. Billions of dollars have been invested in curing you and inventing new technology stacks to cure all your subtypes. In contrast, it's not a great time to be a whale, or a pine tree, or any biological resident of East Palestine, OH.  

Why? Because the better the economic case, the easier it is to mobilize human capital. In medicine, this has taken us to a strange place where dozens of companies work to solve the same, “commercially-viable” diseases1. Too many amazing people work on problems with high professional gravity, rather than truly impactful solutions. It’s great that many deadly diseases are now curable, and that a vaccine for the novel coronavirus was prototyped within days and distributed in months. But human resources are still not allocated toward the most important problems in biotech. This paradox is what prompted Peter Thiel to lament, one decade ago, "we were promised flying cars, instead we got 140 characters."

Similarly, in bioengineering, we were promised solarpunk2 but instead got monoclonal antibodies. We wanted de novo proteins to save us from forever chemicals, but we got antibody-drug conjugates. We wanted fuels grown from algae and instead we barely get pharmaceuticals grown from algae while 98% of our chemical feedstocks still come from oil. And we wanted cell-free manufacturing, but we’re still stuck at laundry detergents.

So how do we work on the most important problems3? There's no shortage of government and private capital looking to boost climate tech, and there's no shortage of talented bioengineers wanting to work on big climate issues, so clearly there is a disconnect somewhere in the middle. We need to increase connectivity between problems and talents in order for biotech to reach its trillion-dollar potential on a timeline that matters for planetary health. 

Potential in scale and precision; Challenge to engineer and deploy.

Biotech is ideally suited to solving planetary-scale challenges because of its two extremes: it is atomically precise and infinitely scalable4. But there are several reasons that biology is not yet harnessed fully to address climate change. 

For one, biology is hard to engineer5. It takes many shots on goal to accomplish much of anything. And when something does work in the lab – usually after burning through lots of time and money6 – the deployment journey is so long and capital-intensive that you’d better hope the initial result is tied to a translational path that matters. 

AI may well be an accelerant, but bio science and bio deployment are still difficult in different ways. 

Beyond just “Life finds a way” to do what it wants, there are simply not yet many playbooks for successfully deploying biotech commercially. If you work with mammalian cells, you’re probably headed to clinical trials with a biopharma giant. If you’re engineering plants, you probably want to sign a partnership with the Big Four7 agricultural oligopoly. These trodden paths to impact are interpretable by capital markets, priced accordingly and staffed organically.  

The way to ignite more transformative impacts in biotech, then, is to build a broader set of platforms to shepherd early ideas into world-changing impacts. Innovation is needed in how we support early translational science and in how we scale those ideas in absence of established playbooks. Such meta-innovation is essential because the existing drivers for biotech today do not properly incentivize efforts that leverage the strongest features of biology. 

Let’s dream about futures that lay on the other side of both engineering and deployment challenges. It’s easy to look backwards and critique, it’s harder to look forwards and lead. Starting from features of atomic precision and infinite scalability, I’ll briefly skim over a few areas that are under-appreciated relative to impact. We will explore more in the months ahead.

Four important bio-based topics that are hard in both science and deployment

Cell-free manufacturing can use the atomic precision of biology across industrial applications. While there are significant practical challenges, there is no theoretical reason why multi-step chemical processes couldn’t be done at scale by self-assembled molecular conveyor belts. Let's tangibly imagine modular solutions of cell-free manufacturing to existing industrial problems: could we one day shrink those enormous factories into sets of programmable flow cells that we assemble like LEGOs? One very likely outcome is that proteins (which alone don’t scale that well due to lifespan) will need to be combined with synthetic approaches (see: Cascade, Caravel, Saipem, Solugen and Plexymer), unlocking chemical frontiers beyond today’s commodities. Just as an airplane doesn’t have feathers, it’s very possible that the way we deploy solutions from nature will look very different from the original context.

Precision tools could harness the self-scaling aspects of biology safely. Very few labs seriously consider experiments of open deployment systems of genetically engineered organisms. This is understandable: An experiment gone terribly wrong could be catastrophic and so the topic is taboo. Is this fear holding us back? What if there was such a thing as public roadmaps for open deployments that weighed the highest impact goals against risk, all framed as a collective engineering challenge? It is important to build at the frontiers of “Precision ecological medicine,” as default paths are leading to ecosystem collapse, and some geoengineering plans like monocrop trees for carbon capture are flawed yet need more than reductive critique. The future bioengineer’s toolkit must harness biology’s magnificent ability to self-organize from the nanometer to the kilometer.

Pathologies caused by environmental issues might be the link between human health and planetary health. It’s easy to ignore melting glaciers or dying species far away, but the human health effects of industrial byproducts are now as proximate as a coal miner’s canary. Cancer rates are increasing rapidly, especially in young people. So too are the rates of autoimmune diseases, infertility and autism disorders. After the best genetic studies have been done on autism, the only conclusion is that genetics alone are little explanation. Then there is the insect collapse and hermaphroditic polar bears.  These pan-species disasters indicate some external factors in the environment, yet systematic monitoring is hard: tiny, mostly unregistered molecules interacting in unknown combinations have unknown biological effects in unknown combinations in unmeasurable situations. But we cannot underestimate the power of an activated, well-resourced community effort taking a data-first, molecules-up viewpoint: if we understand the basis what is currently called “sporadic” diseases, we can start to better price the cost of pollution and incentivize cleaner technology.

Applied Geobiology becomes the bedrock of a new bioengineering boom era. Organisms isolate specific metal ions and dissolve/construct minerals with high efficiency across the planet. This means geobiology is foundational to our future efforts in mining metals, capturing carbon, and harnessing energy. Despite this potential, geobiology has been historically neglected, or at best, its efforts siloed inside industrial practices: In contrast to medical biotech, geobiologists do not have a dedicated suite of tools for technology development, communication, and data centralization. Geobiology has no Cell Atlases, "Kendall Squares" or "Silicon Valleys.” It would be complementary to government funding efforts to grow connective tissue across siloed projects and lay the foundation for an organically growing public/private community.

Community as a bottleneck

ClimateTech is unique because it is a theme of technology growing under external time pressure. It is a novel challenge for humanity to suddenly deploy a trillion dollars in the face of both science risk and deployment risk. 

To make a strawman, it cannot just be a trillion dollars of venture capital driving climate tech. To illustrate this,  just imagine if CRISPR was discovered in 2011 at a startup, rather than a university, and was never published! Even if there was a venture investor comfortable with the science risk of studying bacterial immune systems, the startup would have been alone in solving both genetic editing and delivery problems. It probably would have failed. Then the CRISPR IP would have been acquired by a large company, and locked away; the Innovator’s Dilemma limits the giants from trying anything too transformational. The billion-dollar gene editing ecosystem could have taken decades longer to emerge, if ever.

But what happened instead, fortunately, is that CRISPR discovery was published, the plasmids were made available on Addgene, and gene-editing boomed from 2012-2022. Billions of dollars of value was created and a whole ecosystem of specialists was born. CRISPR, in the sense of its original Cas9 instantiation, is now the kernel of a much broader ecosystem of modern biotech startups.

And, to make the other strawman, it cannot just be top-down programs that distribute all the capital, such as from governments: While roadmapping-based innovation is essential, the ecosystem growth must be organic and cannot be prescribed because the future of science cannot be predicted.8

Let climate biotech refer to the field of research and practice of bioengineering toward a thriving biosphere. Climate biotech must grow organically like the gene editing ecosystem, but faster and across more verticals. And the climate biotech community must grow like the machine learning community has around benchmarks and challenges, but in atoms, not bits.  Problems must be surfaced, platforms created to accelerate efforts toward problems, and pathways to scale must be developed: at the end of the day, it has to be practitioners leading the way.

So curing rare cancers in mice may be a focal point today but the economic gravity has to shift by 20309. Imagine if it was actually a great time in biotech to be a whale inhaling heavy metals, or a pine tree fighting a beetle infestation, or a person in Ohio recovering from unknown chemicals spilled on him. What if you were that person? “Yes that was a dark spot,” you’d think, “but wow are we lucky to have the best people in the world working on our problems.”

Acknowledgements:

In no particular order, huge thanks to Alex Rosay at Cascade Biocatalysts, Cecilia Martinez-Gomez at UC Berkeley, Jenna Hua at Million Marker, Trevor Nicks at Caravel, David Ortega at Phase Biolabs, Henry Lee at Cultivarium, for their technical input and conversation. 

Thanks to Niko McCarty for editing and, well, teaching me how to write.

Thank you to Miranda Stahn for technical and writing edits.

Special thanks to my brother (and Homeworld co-founder) Paul Reginato for being one of the deepest thinkers I know and for using his artist’s soul to always drive to beauty in both art and science. I appreciate his collaboration and support in using this blog space to prototype ideas. 

Haven’t you wished all carbon in the atmosphere would just … go away? Well we think it’s time that Fusion was just solved. Because somebody else can’t make the world a better place better than we are making the world a better place. We gotta move!

It’s time to gather the dreamers with the dreamers, and leave the doers out of the conversation until they start nodding yes. Why can’t our phonovoltaics get more than 90% efficiency? Does gravity really have to apply to all molecules equally? Maybe Desktop Black Holes are the waste solution we’ve been ignoring? If metal mining is so hard why aren’t we dragging gigantic magnets through the ocean? Time to swing the Occam Sledgehammer!

This paradigm-busting weekend will take place over two days in Asilomar (“AWE-SOME-LAR”) in December 2023. Keynotes will be by DERPA-G and a pre-recorded talk by George Church.

LARPA-ing Schedule

Friday night: Welcome Dinner featuring Keto-Vegan Cheat Day cuisine.

Saturday day: 

• Peer-to-peer TED Talks. 

• Fusion Hackathon: 24 hours to get to positive energy production using technology.

• DIY BSL4 Biotech demonstrations and its applications to preventing lightning strikes.

• MetaMetaScience Panel Discussion: Is it “$1B” or “$1BB”?

• Talks:

  • “0 PPM CO₂ and Beyond”

    

  • “One Fuel Rod Per Child: An Abundance Politics Solution to Expand Beyond the 20kg Global Supply of Tritium”

  • “The Inverse Combustion Reaction: The Unreasonable Power of Sign Changes.”

  • “We Found The Nail! How our SaaS-only VC Fund Will Accomplish Global Ocean Alkalinity Enhancement to pH 9 by 2029.”

    

Saturday entertainment: 

• Wikipedia Race Tournament. 1st Place Prize is a $1M donation to a Wikipedia page of the winner’s choice. 2nd place prize is a $1MM donation to the same page.

Saturday night: Whale BBQ, a methane-free alternative to beef. Hosted by DERPA-G (see below).

Screening of the pilot for the greatest show hard-sci-fi thriller that never was, Heat Vision and Jack, viewed in glasses-free 3D.

Sunday morning: 

• Concierge WakeUps Inc. demonstrates how to get 10% more billable time from scientists.

• Workshop on Brute Force Optimism: Hey, There Can Only Be a Finite Number of Wrong Answers, Right?

  

Sunday: 

• Fusion Hackathon Demo Day (geiger counters provided). Winner receives an antique Nintendo GameBoy.

• Fastest Grants: A title-only granting program, presented by an AI rendition of Tyler Cowen’s voice.

• MetaMetaScience panel: “Extending Warren Weaver in 2023: Is Sigourney Weaver close enough?”

  

• BioPlenaries:

  • “(p)Uper: Rethinking the Human as a Mobile Bioreactor” 

  • “BB-SRM (Bovine-based Solar Radiation Management): Ultra High Throughput CRISPR Screens Toward White Cattle.” 

  • “Deployment of Next-Gen Plant Engineering Toward Space Elevators.”

  • “On-Demand Aligned Mitosis (O-DAM): A Just-in-Time Compilation Strategy for Asexual Adult Human Reproduction.”

  • “Baking Cookies with ThermoFisher Reagents: How Going Orders of Magnitude Up the Cost Curve Informs How to Go Orders of Magnitude Down.”

  • “Helping Nature Help Itself: Delivering Correlates of Human Consciousness into Ecosystems Through Broad-Host Engineered Phages”

    

• Closing panel of panels: 

  • “20/23 Hindscience: Judging Scientists of Old By Things They Should Have Known Today.”

    

     

  • “What did Monty Python get Right and What Is Still Unclear?”

  • “FlatEarthing For Good: Empowering freedom of thought by going orthogonal to legacy knowledge”

    

Parting entertainment: Heat Vision and Jack Photo Booth scientific-certification-2023.gif

Co-organizer: DERPA-G

Department of Engineering Really Performative Actions - Geoengineering (DERPA-G). DERPA-G will report on their 2022 summer summit from an Arctic mountain in which they powered a 700-person elite conference using purpose-built giant natural gas reactors. After $100M/MM investment raised via SPAC, DERPA-G successfully captured 69% of the carbon using pilot modular carbon capture. By 2025 they intend to develop the capacity to store such captured carbon, and for this year's summit they will release the captured CO2 back into the atmosphere in a system termed Catch-And-Release DAC (CARDiAC). Attendees were amazed at the quality of the beef tartar that comes from Australian grass-fed cattle with an additional dietary supplement that cuts methane by 44%. DERPA-G additionally reports on the accidental success of whale tartar as a replacement to beef due the excess hemoglobin and the methane-guilt-free keto snack.

How you can try to get an invite

We will put a public application link in compliance with all relevant regulations pursuant to LARPA’s pending hybrid profit/non-profit/nationstate status.

Pre-applications will be posted shortly and we half-heartedly encourage you to apply.

Happy April 1, 2023.

Acknowledgements

This is artisanal, small-batch nonsense: Not a single LLM was sullied with the creation of this gibberish. Almost all early readers asked if GPT was involved: this bodes poorly for my comedy career. If you have ever seriously pitched any of these ideas, I promise you the coincidence is accidental.

The original humor idea came from a running joke with Paul Reginato (Homeworld Collective) and Kristin Ellis + Ryan Orbuch (Lower Carbon Capital) last week at a conference. Also thank you to Michael West (WestComms) for his additions.

This post is copied from an email Q&A with Tom Kalil, Chief Innovation Officer at Schmidt Futures. Tom is the [Q] and I’m the [A]. This conversation originally started around my fascination with the progress in Program Synthesis and its future influence on how we do biological research. Every day we see a new demonstration of the power of large language models, and it seems the future integration with biology is getting rapidly closer.

The joy of publishing material in blog format is that it encourages discussion. These are forward-looking ideas with no guarantee of accuracy beyond me doing my best in early February 2023. I recommend Sam Rodriques’ Tasks and Benchmarks for an AI scientist that was published yesterday which explores adjacent ideas and I think coins the term Large Science Model (LSM). Comments and thoughts welcome.

Thanks to Tom and all others for their contributions and conversations.

Q: Dan, what I wanted to explore with you is why you and other researchers believe we can accelerate the pace of scientific progress by combining three different technologies – namely, generative AI, cloud labs, and in silico prediction.  Why don’t we start off with discussing the individual building blocks – and then we can discuss opportunities for combinatorial innovation.  What is generative AI, what is program synthesis, and why are they relevant to science?

A: The big picture here is that the interface between bits (software) and atoms (physical world) is expanding at an astonishing pace. Most of our Q&A will focus on the bits→atoms direction (eg, software controlling physical experiments) and specifically I think Program Synthesis (roughly, '“code that writes code”) has significant potential to accelerate progress. But the story starts in the atoms→bits direction.

The massive data created in biology has been the resource for powerful computational tools. Most biologists have probably seen the NIH chart showing that sequencing data outpaces Moore’s law. However, in my opinion, the best metaphor for this new era of biological data is that a single optical microscope in a modern biology lab, if run 40 hours per week by a single scientist, can produce more raw data annually than the Large Hadron Collider (once the poster child of team science in the era of Big Data). Paul and I learned this first-hand in our PhDs! Bigger datasets mean bigger opportunities for computer science, and the resulting software breakthroughs have increasing implications in the physical world. 

More than “just” processing data from biological experiments, we’re entering the era of AI-powered experimentation. Recently I wrote an essay called AI is Part of Biology’s Future to explore the inevitable 3-way intersection of three major technologies:

• Generative AI: a term for leveraging publicly available Large Language Models (LLMs) like GPT-3 that can power robotic planning and human-level chatbots.

• Cloud Labs: centralized robotized lab facilities, eg Emerald Cloud Lab.

• in silico prediction: exemplified by the protein folding work of AlphaFold and ESM which now publish 100M+ structures.

I believe we will see that the deep integration will be more than the sum of its parts. 

OpenAI’s ChatGPT has been in the world for a few months and was heralded as “AI’s iPhone Moment.” If any readers haven’t yet played with chatGPT, I strongly advise them to do so now while referring to Riley Goodside’s Twitter feed for examples of humorously illuminating expert usage. ChatGPT is a general purpose LLM, and later we will explore how LLMs might change as they ingest more scientific information. But first, let’s start with the most powerful capability of Generative AI: Program Synthesis.  

Program Synthesis is a subfield of AI that suddenly went from academic pursuit to a new foundation of industry. As long as we’ve had software, we’ve had the wild idea that a software agent could solve arbitrarily complicated problems by writing its own programs. I had my “oh wow, this is serious” moment when I sat in the audience of Kevin Ellis’ MIT PhD thesis defense in 2020 and was floored by his DreamCoder work. Despite the beauty of DreamCoder, it was academic in scope and programmers in 2021 scoffed at any practical implications of GPT3. But it turns out that LLMs have unlocked Program Synthesis and we can now see the world-changing potential in the publications from Microsoft, from Google and from OpenAI. And as software reaches into the real world, the implications can be seen in SayCan work from Google and Everyday Robots (video below) which demonstrates complex natural language tasks being broken down into a sequence of robotic actions. 

Even more impressive to me than human language instructions is the new ability to prompt a software agent with just a unit test (paper). A toy example in software would be a function to confirm two numbers are correctly multiplied together and a toy example in biology would be a PCR amplification to detect the presence of  DNA sequence (like a Covid test). Now it’s probably a weekend hackathon for chatGPT or similar to connect to a robotic lab setup to do such a PCR test. These examples may seem trivial, but the takeaway from 2022 is that practitioners move beyond toy problems very quickly. We are in an environment rich with positive feedback. 

Q.  What are some potential advances we could make by training large language models on the entirety of the scientific literature?

A: It’s inevitable that somebody will ingest all of scientific knowledge into a foundation model like a GPT1. Any of the leading groups could probably do this, and probably are already working on it. For example AllenAI has many public tools for processing scientific literature. But it’s hard to directly answer the question of what advances we could expect. An LLM with all scientific knowledge might be shockingly good or it might be deeply disappointing (see the ScholarBERT paper): As I’ve written about recently, we have to embrace the unpredictability ahead. 

Let’s briefly talk about the opportunity of mining all scientific literature with the intent of creating a Large Science Model (LSM). Only 3% of the training data of GPT-3 was Wikipedia, the rest was text from web crawlers and books. The Chinchilla paper (excellent explainer), underscored the importance of finding new, high quality information to ingest into these powerful compute models. So compare GPT-3’s training data of 3 billion tokens (tokens are how the LLMs see words in text) from all of Wikipedia to SciHub’s growing database of 88 million papers, which my back-of-the-envelope math tells me there is an untapped resource of about 200 billion tokens. (ScholarBERT raises good questions on how to make use of the data). This certainly appears like potential for a big improvement, especially since there are now scientific question and answer benchmarks. So, if somebody navigates the legal challenges of ingesting all of SciHub and sharing the new LSM with the world, there might be a really powerful platform on the other side.

So let’s assume there is a publicly accessible, LSM, what might we do with it? As discussed earlier regarding Program Synthesis, one very powerful feature of LLMs is that they can serve planning functions. Perhaps the next-gen science LLM could auto-draft roadmaps by breaking down big problems into tractable projects or predict next projects (eg, DELPHI). Furthermore, the science LLM could execute day-to-day scientific research in the lab if it were paired with domain-expert AI sub-systems (eg, see Amyris’ Lila for microbial engineering). This dynamic parallels a human team: one manager to scope the project and a set of subordinates to plan, execute then report back the experimental results with suggested next steps. One advantage is the inherent scalability of automation, and the bits→atoms bandwidth could widen dramatically with cloud labs.

Q:  What are cloud labs?

A: Named in reference to cloud computing, cloud labs are centralized robotic facilities for conducting life science research. Currently, there are a few commercial players like Emerald Cloud Lab, Strateos and Culture Biosciences, and it will be interesting to see how this field grows. In the big picture, I’m optimistic that bioautomation will be widely adopted, but we have to appreciate the short-term uncertainty. On one hand, the centralization of expensive hardware and orchestration of logistics (reagents, storage, data, etc) seems to be directionally correct due to economies of scale. But on the other hand, there is currently significant economic and technical friction in the transfer of a local lab to a cloud lab. 

This is where Erika DeBenedictis, founder of the BioAutomation Challenge and an innovator in synbio+automation herself, is such an important thought leader, and I’d point readers to her recent article on how cloud labs can better integrate with academic research.  

Q.  What is in silico prediction, and why are scientists so excited about progress in fields like protein structure prediction?

A: I say in silico prediction to refer to the task of estimating molecular properties. These properties could be 3D shape, reactivity, stability, enzyme kinetics, substrate affinity or many other such characteristics. Although this might sound mundane at first, it’s important to understand the implications. If you solve the challenges of molecular-scale prediction, you unlock the ability to virtually test and iterate millions of times faster than could ever be done in reality. The frontiers of in silico prediction touch everything from solar panels and batteries to medicines and fertilizers. 

The most visible example of in silico prediction today is protein folding. It’s brilliant that we can draw a 3D protein structure directly from a 1D DNA sequence. Google’s AlphaFold which has now published 200 Million protein structures, and Meta’s ESM Atlas has published 600 Million protein structures. This changes how we mine the troves of biological data: we can now search directly for 3D and functional characteristics, which is useful when scanning nature’s toolkit for something like signaling peptides to control animal or plant biology. Or if you’d allow to me wade into weird topics in biology, I’d love us to have a better understanding of how (and if) microbes can eat radiation: could we find photosynthesis-like protein systems in the genome of those weird creatures that live on spent uranium fuel rods or the outside of space stations - do they actually harness energy in a totally novel way? These new computational tools can help us explore difficult-to-study biology, probe early ideas quickly and also unlock entirely new directions of bioengineering.

We’re now in a rapid growth period of de novo protein design in which we think of proteins as programmable nanoscale machines. One direction that is particularly exciting to me is the inverse protein folding problem: Let’s say you know a protein’s shape and function, but you want it to be more stable than the original to be used industrially, or have more human-like sequences to minimize immune response in medicine. Or, if you’ll allow me to be speculative again, one of my technical fascinations is the frontiers of artificial enzymes, in which we discover a tool in nature but then replicate it in a scalable, synthetic format (“airplanes don’t have feathers” is the classic quip of learning from nature but scaling it up in our own way). Inverse protein folding is the category of problem that would solve these challenges: Such difficult tasks are becoming more tractable every day.

I have to channel my heroes of experimental biology to give the caveat that these computational tools just give *predictions* and it all needs to be verified. We can’t fold a few proteins and think we’ve solved biology. These in silico prediction tools can and will fail catastrophically, plus their training data might be missing critical knowledge like post-translational modifications, phase-separated nano environments or self-assembled structures in vivo. Such missing data is critical for climate-relevant applications like Methane oxidation to mitigate the greenhouse effect or the “nitrogen-splitting anvil” to develop future fertilizers.  So it’s excellent that we can generate so many candidate molecules but we have to remember that it is a long pipeline toward deployment, and this is where robotics has a large role to play. 

Q.  How might we accelerate the pace of scientific research by co-designing these three technologies?

A: There is the potential for three trends to reinforce each other: AI that is getting good at planning and delegating tasks, AI that is getting good at molecular-scale computation and robots that are getting better at doing physical work. Together, this is potential for a virtuous cycle of more physical science tasks being accomplished by more lab robots that are controlled by a scientific planner AI that then creates more and better tasks. 

This is a special moment in time when we currently stand on the flat part of the exponential curve and can ask ourselves the hard questions: How can we steer these new technologies to work on important planet-scale problems? What is needed to kickstart this flywheel for problems that might not yet have market forces to incentivize activity?

Each technical discipline has a different set of needs and motivations, so it does require some co-design: Computer scientists need datasets, biologists need assays, roboticists need tasks. In the diagram below, I briefly sketch out how the full virtuous cycle could be viewed as discipline pairs: it’s possible that it might be initially productive to consider specific, representative challenges along each edge. For example, what would it take for teams of LLM researchers and biolab automation experts, who probably don’t yet intersect much, to collaboratively create a set of challenges and data that inspires both of their fields? The role of common resources cannot be overstated as the role of public datasets has been foundational to machine learning’s progress. As we collectively ponder about how to get each discipline the resources they need, we can also identify application spaces that might inspire the communities to self-organize into projects. 

It’s important to consider which subfields of biology will get access to these AI+robotics capabilities without interventions today: it is relatively straightforward to write business plans that justify bringing automation to certain aspects of medical biotech (eg, antibodies). But what about high-impact, high-risk areas like green hydrogen (high CapEx, low margin), carbon farming (no market forces yet) or curing diseases caused by pollution (hard challenges in both science and business) ? These example topics could all be billion-dollar opportunities with positive externalities, but because they have both market and science risk, they are currently the last fields to benefit from cutting-edge AI+robotics automation. 

We are building Homeworld Collective to grow the community of climate biotech research and to make sure climate problems are well-framed to make the most of cutting-edge technologies.

We have major technological challenges to ensure Earth can support a thriving biosphere for the centuries ahead, and AI-powered experimentation could give us the speed we need. We’ve mentioned a few climate-relevant topics so far – methane oxidation, nitrogen fixation and radiosynthesis– and there are many more that could be significantly accelerated with by AI-powered experimentation: next-gen metal mining, decentralized energy production, cheap energy storage, utilization of waste biomass, replacing petrochemical feedstocks, biological resilience, rapid topsoil recovery, etc.. But each one of these are big problem spaces that warrant their own white paper to expose the most actionable sub-problems ready for AI-powered experimentation. It’ll be many years, if ever, that we could use the toolset and enter into the chat box “solve carbon capture,” I’ve tried it!

Q.  What’s a plausible scenario for how this might play out in a particular application of science, like drug discovery or enzyme design?

A: Personally, I love the topic of understanding and enhancing biological resilience. I think now could be its time. 

One specific subproblem inside resilience that could benefit from AI-powered experimentation is how to make plants stronger and more robust against environmental threats. I see a lot of work for plant engineering that is genetic in nature, but I think we need solutions that are faster to develop and more controllable to deploy. Said in a funny way, I’m curious about the frontiers of drugging plants. 

First, why would we drug plants? People have long engineered small molecules for crop protection (pesticides etc), but what about for specific climate applications like carbon capture, marginal land repopulation, metal uptake or photosynthesis overdrive (irrespective of food yield)? This is obviously speculative but an interesting direction of thought because small molecules have the potential to go to market much faster and cheaper than a genetic modification.

Developing and delivering small molecule modulators for given protein targets is a well-established playbook in human medicine, so perhaps this process could be replicated broadly and rapidly from the ground-up with AI-powered experimentation. If we could invent small molecules that confer resilience against heat, drought or light stress, then we might have a just-in-time solution to save crops in extreme weather.

There’s enough here to bring AI scientists, plant biologists and roboticists to the table. The plant literature has plenty of metabolic networks for the future scientific LLMs to mine, next-gen robotics companies like Hippo Harvest show us that robots can work with plants at scale, and in silico predictions have shown us that we can screen millions of molecular variants to impact a given pathway. 

Let’s say we first build the robotic lab setup, document the assays and share the output data to the world (images and ‘omics of each experimental condition across multiple parts in the plant). Then we make some initial data (ie, positive and negative controls, indicate the ability to scale up to more test molecules (say, using existing drug libraries) then create publicly accessible datasets from the whole system. This creates public goods in the form of knowledge and lowered barriers to entries for experts to engage with the problem.

This is thin on details for brevity, but hopefully illustrative of an opportunity to create assays for biologists, tasks for roboticists and large datasets for computer scientists, while also translating known recipes for success from medical biotech.

Q.  How does this relate to your vision of enabling “hyper-productive” fields – like the role that open source software and cloud computing played in lowering the barriers to Internet entrepreneurship?

A: Everything we’ve discussed so far is about technology, but it is ultimately human beings working in groups that turn potential into impact. I want to see more research in how we build and nurture new research communities2. 

Machine learning and medical biotech are gold-standard research communities that have produced several world-changing leaps in our lifetimes. These communities have common social, logistical and economic structures that create windows of hyper-productivity, and I believe it may be possible to engineer hyper-productivity in the future.. 

As an illustrative and famous example, when the AlexNet paper from Geoff Hinton’s lab became the foundation of the Deep Learning Revolution in 2012, it was a leap developed on a large infrastructure of open source tools and community knowledge, notably centered around Fei-Fei Li’s ImageNet dataset. So while it may be useful for a resource allocator to ask “how do I find more Geoff Hintons?” it may be even more productive to explore to ask “how do I recreate the computer vision community from 2010-2015?”  

So in this moment, when there is technological momentum and major global challenges in need of fresh ideas, how do we plant the seeds for hyper-productivity into AI-powered experimentation?  My current mental model is that there are six conditions to engineer, and I’ll give brief examples of each one:

• Common platform for experiments: Open source computational libraries like Torch/Tensorflow, SciPy and Pandas have been essential in the deep learning revolution by making it very easy for a practitioner to be conducting research on the same toolkits as the state-of-the-art methods.

• Common platforms for scale-up: Cloud computing means that if you build an app that works for a thousand people, it can work for a million people.

• Common goals: In biology we had the Human Genome Project and the CASP protein folding prediction challenges.In computer science, public challenges like ImageNet have been instrumental in focusing many efforts in an apples-to-apples comparison. 

• Enforceable honesty: A best practice in computer science is to share your code and models so people can replicate your results for themselves. In synthetic biology, people share their plasmids on AddGene.

• Playbook for outsized success: Y Combinator became the preeminent startup accelerator with a storied history of 20-somethings becoming multimillionaires by following common strategies. 

• Funding to explore: Doing software work has been historically cheap, the semi-serious joke in Silicon Valley is you just need enough money to keep people fed on instant ramen. Bio research is more expensive but there has been lots of available capital 

For AI-powered experimentation, I would suggest we start with collaboratively developing common goals to motivate practitioners and developing common platforms for experimentation to increase the accessibility to the highest leverage problems.

The views and opinions expressed in this blog are the author’s own and do not necessarily reflect the view of Schmidt Futures.

I was recently in a room in which the great economist Tyler Cowen led a conversation around the fascinating question “Is the world getting harder to predict?” This provocation stuck in my head: When one looks at the unexpected Ukraine/Taiwan/EU global political dramas, Twitter’s cowboy takeover or FTX’s explosion into the biggest outright fraud since Enron, the world certainly appears more volatile. So a knee-jerk answer to Tyler’s question might be “yes, of course the world is getting harder to predict!” However, I’ll push the uncomfortable point that more and more of our world is getting shockingly predictable.

As citizens of the hyperconnected 21st century, we must keep in mind that advertising has been a core financial engine of growth. This means that a trillion-dollar effort1 tracks and predicts everything about you. We take for granted how good our Amazon suggested purchases are, how sloppy our Google queries can be, or how immediately interesting our Instagram/Twitter feeds are when we log on. We see OpenAI’s outpainting feature (shown below) demonstrate the expanding neighborhood of predictability2 around any human effort. Similarly, the eerily good AI-generated Joe Rogan and Steve Jobs podcast shows us that beyond predicting what you might say, software can now predict how you might say it. Put simply, if you are easily modeled, then you are easily predicted3, and this will have increasing implications in the future.

So this seems there might be a paradox: how can more and more of the world be increasingly predictable while the biggest parts are increasingly unpredictable?

At first I thought there might be strange interactions across magnitudes. “Small-scale increasing predictability” combined with “large-scale decreasing predictability” has a flavor similar to the Turing Pattern; differential equations posed by Alan Turing that capture short-range and long-range inverse effects (see the pufferfish below). But this initial thought is unnecessarily complex and unlikely to be insightful. However, this initial framing did help uncover a critical assumption: this mathematical model makes the strong assumption that the big pieces of the world are proportionally driven by the predictable smaller parts. I don’t think this is a correct assumption, which was the insight into a much simpler answer.

The argument

Question: “Is the world getting harder to predict?”
Answer: “Yes, because the hardest aspects to model are becoming more influential.”4

Here is my argument with foundational books to each idea:

1. Increasing technology means key decisions have increasing reach (David Deutsch's "Beginning of Infinity", 2011).
   “Beginning of Infinity” is one of my top recommendations for anybody embarking on a scientific journey. Deutsch’s big idea is that Explanatory Knowledge (according to the glossary, an explanation is a “Statement about what is there, what it does, and how and why”) necessarily creates more Explanatory Knowledge. This positive feedback creates an exponential growth of explanatory (ie, useful) knowledge which can continue to infinity, providing certain criteria are met. We are on the steep end of the curve on many dimensions of technology today (eg, “AI” applicability and synthetic biology) in which decisions have global implications.
   

2. Key decisions are being made by individuals empowered by new technology. (James Davidson and Lord William Rees-Mogg's "Sovereign Individual", 1997).
   If Beginning of Infinity is the techno-optimist manifesto, then the Sovereign Individual might be the techno-pessimist Nostrodamus’ Prophecies. The core idea is that modern democracy emerged with the Industrial Revolution because new technology acted as an equalizer of workers' outputs: the assembly line ensured that each worker creates a similar amount of value. Davidson and Rees-Mogg argue that the software revolution will have an undoing effect on social equality: software will accentuate the differences of individual ability which will then decrease the validity of democratic governance. This was impressively prescient given that it was written five years before broadband wifi and a decade before the iPhone or Bitcoin. Davidson and Rees-Mogg would see the powerful founder+CEO+celebrity archetype of the 2000s as corroboration to their thesis: technology now makes some individuals as powerful as nation states.
   

3. Large groups actions are easy to predict and decisions made by representative members of large groups are easy to predict. But decisions made by high-agency, non-representative single individuals are very difficult to predict (Asimov's Foundation Series, 1951).
   Asimov foresaw the world of Big Data in the 1950s when he built a SciFi universe around Psychohistory, a fictional branch of mathematics that sees the future by leveraging the fact that humans en masse are highly predictable. Such a concept comes from physics: the overall behavior of a gas follows a few physical laws, but a single molecule’s path is entirely unpredictable. The Foundation Series tells the story of a single individual who solves the equations that predicts a thousand years of interstellar humanity5. Even more prescient, the Psychohistory prophecy has limits and foresees windows of time in which an individual will emerge who determines the outcome of a crisis. But who that person is, and whether their actions keep humanity on the foreseen path, is unknowable.

4. Therefore the world is getting harder to predict because decisions made by a few key individuals are of increasing consequence.

Examples of modern unpredictability

History is punctuated by legends of single individuals. Hegel coined the term “world-historical individual” in the 1800s to describe people like Napoleon and Caesar who have an outsized mark on humanity’s timeline. While the historical kings and generals that Hegel studied were certainly unpredictable, the modern world-historical individual feels a difference-of-kind apart because the modern individual can build her own platform of global reach. A few pertinent examples of technology-powered individuals changing the world today:

• A single 30-year-old individual, Sam Bankman-Fried makes a $22 Billion company that vaporizes overnight in spectacular fashion. Some say they saw the red flags in early 2022, but there were multiple years of operation that were so fraudulent that the top-tier bankruptcy team says FTX is the worst case of “mismanagement” they have ever seen. It’s more than just financial fraud that makes this case interesting: Before the blowup, “SBF” very publicly aligned himself with forward-focused Effective Altruism philanthropy movement, and regardless of SBF’s authenticity, he substantially accelerated the global EA growth.

• Elon Musk takes over Twitter in either an erratic or brilliant manner, depending on whose opinion you get. As a private company, Twitter is now free from public shareholders demands for conventional leadership. The outcome of this saga is unpredictable because it is entirely dependent on Musk himself. On one hand, scores of advertisers are leaving with dozens of labor lawsuits following the 80+% reduction in force. On the other hand, it’s possible the remaining Twitter team performs excellently, Elon brings more activity to the site that is still the global town square, and the newfound agility positions Twitter uniquely well for a landmark event such as TikTok being banned in the US.

• Donald Trump’s becomes President of the United States. Aside from the Bio-SciFi book RiboFunk (1996), I don’t think anybody could have predicted Donald Trump becoming the political force that he became. But that’s partly because very few people predicted how integral social media became (maybe Clay Shirky and tech executives saw it coming). Regardless of how you personally feel about Trump, his mastery of social technology must be appreciated. Smart friends are now pitching me on Kim Kardashian and Jake Paul as future politicians: I don’t have an opinion here but do think it’s easy to underestimate the intellect of social media stars who can single-handedly build empires the size of heyday cable TV channels.

• China is run by Xi, Russia is run by Putin. Two major countries on the global stage appear to be entirely controlled by single individuals, of which technology-powered state control is an essential piece of their strategies. There seems to be increasing tensions between the populations and the governments in these countries, but the outcomes are totally unpredictable.

  

I think some might try to ascribe a value judgment to this increasing unpredictability. My perspective is bluntly that we can’t go backwards and any attempts to do so will only backfire6: Macro unpredictability is here to stay as long as technology develops, and technology will surely continue to develop. If we appreciate the implications of an unpredictable world, I believe we can create much good in the future ahead.

How I learned to stop worrying and love the unpredictability

What are some ways to think about thriving in a world that is made of many easily predictable elements and a few main drivers of variance? We can take a quick look backward before looking forward, and the history of physical science is a great case study.

We take for granted that the science we use today is built on a totally unpredictable chain of unique individuals and chance events. There’s been increasingly more study of this phenomena: Nature posted a blog about the unpredictability of scientific success, and I love reading the Metascience/Progress Studies work7 such as New Things Under The Sun, New Science, or Ben Reinhardt’s and Nadia’s Asparouhova’s blogs. No NSF officer could predict the impact of an individual genius on the level of John Von Neumann, although Warren Weaver’s ability to cultivate Nobel laureates may be the current record. We can’t manufacture genius (yet?), but we can prioritize development of individual agency and aim to build the conditions that promote scenius (magical pinpoints of group creativity).

It is surprising that bicycle makers were the first to take flight, but the Wright Brothers moment of triumph is a tale of iteration and hyperfixation that reflected their agency. They chose the problem of tinkering with flight because they loved it, and they won the race to flight. But what about explorations that aren’t as clearly defined?

Much progress in science (aka, resolutions to Kuhnian Crises) was unpredictable and, as such, initially rejected by contemporaries. The story of the birth of thermodynamics is particularly illustrative8 : Sadi Carnot was a French mathematician who self-published a failed book that printed 100 copies. He received little attention in his life for his ideas and he died poor, alone, and insane. Two years after his death, a French professor found a copy of the book and wrote about it, which made its way to James Prescott Joule, a beer brewer’s son in England, who tinkered to experimentally explore the world with brother. In 1847, Joule applied to give a talk on his explorations of Carnot’s ideas to chemists at the Royal Society meeting but was rejected for being too strange, he was instead given just a small time slot to read a summary with no discourse allowed. But 22-year-old William Thompson was in the audience and spoke out of turn because he was so impressed by Joule's ideas, and a lively discussion ensued in the session. Years of collaboration later, Thompson became Lord Kelvin, Joule is immortalized in our concept of energy, and Carnot is forever known as "the father of thermodynamics." Stories like the chance encounter between Joule and Thompson can be cherrypicked throughout the centuries, but now the impacts of such events now can happen in weeks, not decades9.

In addition to celebrating the hidden gems of unpredictability, we must also appreciate the failure modes of forcing predictability. In an excellent Q&A, Tom Kalil references the book Seeing Like a State by James Scott: this book is an essential list of high-budget failures of governments forcing complex systems to become manageable. Examples include natural forests in post-feudal Europe that were converted into unstable monocrops which eventually collapsed, and cities planned from airplanes by foreign architects (eg, Le Corbusier and Brasilia) that resulted in barren spaces that were used in opposite to the best-laid plans. The highlight of Scott’s book was a narrative of Jane Jacobs’ successful fights against the top-down planners: while Robert Moses and Le Corbusier were flying in planes, Jane Jacobs was walking at street level observing how people lived. While every architecture student knows her for alley-level view on city planning, I think her brilliance extends into technical progress too. Jacobs used the metaphor that rigid army formations only makes sense in parades for managers and an army’s functional form would look like chaos to anybody but the practitioner. If we’re trying to optimize for creative output in positive direction the decades ahead, we need to consider this idea of a functional form versus a manageable form. This will look like cultivating agency in individuals by giving opportunities to explore and a careful reimagining of what managerial metrics actually matter.

Concluding with some ideas in motion

If the world is to become more and more unpredictable, then the predictable parts are in danger of having less and less impact. If we appreciate the magnitude of how much technology has developed (go play with ChatGPT if you haven’t yet) then we must act with urgency to adapt our systems and institutions, especially for personal development (aka education) and scientific research.

Though I write with optimism now, I admittedly felt an initial unease and a desire to cling to a more predictable existence that I’ve always known. Everything is moving faster: we used to think of cultural generations as lasting a decade, but now I think it's closer to two years between cultural shifts. This journey from unease to optimism is similar to my realization of how much of biotech research is going to change with full-stack automation. There can be beauty and fantastic outcomes in the unknown ahead, it requires good people being empowered and engaged.

I’ll conclude with some draft ideas for future exploration:

• Self-agency might be the core prophylactic to predictability. It is more important than ever to develop ourselves and our youth into high-agency individuals. Sir Ken Robinson’s 2006 TED Talk (the most watched TED Talk of all time) was pushing this thesis using the language of enhancing creativity: His take was that schools have to to get out of the business of selling facts and focus on developing creative freedom. Similar to the point in “the Sovereign Individual” that assembly lines equalized economic value per worker, some argue that modern factory schools were designed to produce workers for those assembly lines that don’t exist anymore. These heavy critiques apply to all education levels from kindergarten through the PhD, and my guess is that we’ll see the most changes at the college-age first. If we train predictability and manageability, we may be training replaceability.

• The question is *not* how to decelerate technology’s development, but what interventions we can do that accelerate technology’s growth and impact in areas that might organically be the last recipients. For example, online ads and pharmaceuticals will be the first recipients of any advancements because startup logic drives innovators to high-paying markets. But what about frontier fields like artificial enzymes or historically high-walled gardens like nuclear energy?

• Let’s actively study and experiment how to create conditions that encourage more unpredictable results of individuals or groups in hyperproductive communities.

If you read Martin Scorsese’s classic 2019 roast of the Marvel franchise, you’ll see he pleas for us to again turn toward unpredictability. The 77-year-old master filmmaker received a lot of attention for calling the Avenger movies “theme-park rides”, so he wrote a longer essay to distinguish what he called "audiovisual entertainment" from cinema. In his words, true cinema comes from a single artist's vision, whereas audiovisual entertainment is “market-researched, audience-tested, vetted, modified, revetted and remodified until they’re ready for consumption.” To make the financial return predictable, the content is made predictable. So instead of accepting the risk of an artist’s vision that can last forever upon success10, we get shiny but forgettable content of diminishing value. Now mix into this critique the technological aspect: Stability.ai is now generating photo-realistic images at effectively real-time speed, meaning that the audiovisual entertainment industry is about to get overhauled. One ironic upside is that we may have an Iron Man film made by a one-person team.

Might a one-person Iron Man production actually become cinema again in Scorsese’s eyes? That is unpredictable!

Subscribe now

I hope to continue discussing these ideas online, offline and in future posts. I want to thank Rick for reading and shredding a draft of this essay :) I wish I had time to factor in all his great feedback, and specifically appreciate him pointing out important initial omissions such as Hegel and Shirky.

Credit to OpenAI’s chatGPT for helping me rewrite sentences and paragraphs that were clunky.

Also, the process of thinking and writing this essay motivated me to read Tyler Cowen’s and Daniel Gross’ book Talent.

Edit: Cool to see this linked at MarginalRevolution :)

On April 10, 1964, the piano phenomenon Glenn Gould played his last concert. This is historic because, at the ripe age of 31 (he had played professionally since age 14), he quit public performances to commit his career to recording. All of the money in that era must have been in concerts, especially for classical musicians, so it surely looked like madness to voluntarily leave the biggest concert halls for tiny recording studios. With respect to his commitment to the studio life, Glenn said the following about recording music:

It’s the future for music:
it’s the future of performing of music,
it’s the future of writing music,
it’s the future of listening to music.
All of our futures in music are involved with recording.
It’s as simple as that, and as complicated.

Glenn Gould turned out to be very right. He was so correct that we take for granted that we live in the exact musical futures Gould described, and his vision sounds obvious now.

If a standout talent like Glenn Gould were to say this same passage above in 2022, what domains or trends might that person be referring to? It could be a biologist talking about how we understand and engineer biology in the early years of AI:

It’s the future for life sciences:
it’s the future of doing experiments,
it’s the future of analyzing experiments,
it’s the future of conceptualizing new experiments.
All of our futures in biology are involved with AI.
It’s as simple as that, and as complicated.

Hyperbolic statement or not, the bits-to-atoms interface is accelerating. I'd like to explore three major developments that I’ve observed that seem to be converging. We could visualize the three trends in a rough hierarchy:

I’ll briefly explain each element independently before going into their intersections:

1. Generative AI is the blanket term for large, general purpose neural networks (costing millions of dollars in just compute time to train) that are now publicly usable. Large Language Models (LLMs), exemplified by GPT-3, have surpassed a threshold of capability which takes humanity into a new era of software. Specifically, the field of program synthesis (“code that writes code”) is a powerful new frontier that is in early stages of industrial deployment in computer programming but may soon have deployment applications in other fields.

2. In silico predictions describes molecular modeling efforts that had previously been intractable or relatively stagnant, but have recently have undergone leaps of progress1. Exemplified by the protein folding work of Google’s AlphaFold (200 Million protein structures) and Meta’s ESM (600 Million protein structures), in silico predictions have translated from basic science to drugs currently in clinical trials. Innovations for problems like reaction prediction, docking estimation and inventing molecules of arbitrary 3D shape are driving a new wave of startup formation. 

3. Lab automation are exemplified today by cloud labs, which are a core piece of the inevitable automation coming to biology. In direct reference to the cloud compute services like AWS, cloud labs centralize CapEx-intensive robotic and wet lab resources to expose a programmatic interface for customers. Led by Emerald Cloud Lab, cloud labs are creating publicly-accessible automated services that convert bits to atoms. The thought leadership of BioAutomation, led by Erika DeBenedictis, is catalyzing the researcher community’s adoption of cloud labs in academia and early stage efforts.

What are trajectories we can observe today?

Until there are sizable incentives to do otherwise, cloud labs will continue to prioritize large contracts, likely working on high-margin, low-volume biologics in the context of human medicine2. Today, high upfront cost in terms of time and money on both the provider and customer side hinder broader adoption of general purpose cloud labs. This is why the BioAutomation Challenge is so important: it plays an essential role in accelerating cloud lab adoption in leading biology labs by (1) solving economic hurdles and (2) supporting the production of the first examples of public cloud lab source code. Thanks to Erika et al, I expect us to see dozens to hundreds of protocols in the years ahead, which could catalyze a culture shift that encourages protocol sharing.

In silico predictions are being used to today to seed libraries of molecules into automated pipelines and will continue to grow in adoption within the medical and agricultural industries. These industrial efforts and technologies will be overwhelmingly focused on the constrained problems involving a single protein (eg, antibodies, Cas-like programmable tools, drugging a target, etc.) with known biology3. The deployment will be a human determining the specifications for the software to solve, then CROs, robots or in-house scientists doing the experiments, and finally a human making judgement calls on what candidates to keep or discard. Individual companies may iterate on their processes to be more AI-powered, but I expect those improvements to be bespoke and private until a major software platform breakthrough.

Generative AI is entering a period of high growth of startups deploying foundation models. Sequoia Capital has a good overview of the first wave of growth in the expected areas of B2B services (eg, customer support replacement), graphical design (eg, product advertisements) and, perhaps most powerfully, programming (eg, CoPilot powered by Codex). Program Synthesis, “code that writes code,” is an area of computer science that went from purely academic to suddenly mainstream: Google published utilizing LLMs to program (Austin et al 2021), and Microsoft demonstrated the ability of an AI to write code to given only a unit test to pass (Haluptzok et al 2022). In addition, two other important things are emerging from LLMs:

• Implicit Knowledge rather than Explicit Knowledge Graphs: Computer Science leaders like Jure Lescovec (eg, SMORE paper 2022) have done pioneering work in creating knowledge graphs from various sources of information. What is surprising to many who explore the giant black boxes of LLMs is that it appears there is some sort of knowledge graph that links concepts stored inside the model from training, but we do not yet know how to formalize it or study it. Prompt engineering, led by Riley Goodside and others, seems to be the best approach at the moment.

• Demonstration of mathematical knowledge being represented and manipulated with LLMs: Google’s Minerva project showed the quantitative reasoning ability of language models by ingesting 118GB of data from Arxiv into a pre-trained PaLM model. Similarly, autoformalization is the process of converting natural language mathematics into a formal structure and is now being done with LLMs. And recently a multi-company, multi-institute collaboration made an even more powerful theorem proving approach called Draft, Sketch and Prove. This work shows the ability of the latest neural networks to formalize natural language technical content and to apply (what looks like?) first-principles reasoning.

Given all the activity recently, I offer three theses for discussion:

Thesis 1: The creation of knowledge models may be more important than automation source code examples. Synthace is already building software to abstract the user from the fine details of lab automation code. For human-human knowledge transmission, publicly available cloud lab source code or protocols is essential for reproducing and building upon other’s results. But for an AI agent to determine specifications for a given engineering task, the agent will likely be traversing knowledge in some explicit or latent space and could output results in a formalized language that would be trivial to put into a cloud lab. GPT-3 has been demonstrated to already have capabilities for step-by-step logic and knowledge querying.  This thesis could be explored today by seeing how much extra work on top of an existing foundation model is needed for GPT-3 to give instructions on how to do a PCR covid tests in a lab setting.

Thesis 2: Somebody is going to ingest all 100TB of SciHub into a Foundation Model, it’s just a matter of who, when and what (if any) the subsequent legal ramifications will be. GPT-3 was mostly trained on web crawling and books, so only 3% of the training weight was wikipedia in 2020 (6 million articles, 50 million other pages of content, 3 billion tokens, average of 584 words per article). For reference, SciBERT was trained on 1.14MM papers from SemanticScholar which has 3 billion tokens. SciHub has 88MM papers, and if we assume that we can extrapolate the Semantic Scholar dataset statistics (2600 words per article) with some paper loss due to old/faulty PDFs, it could be reasonable to expect 200+ billion tokens of scientific knowledge, roughly 10x bigger than the Minerva training set of Arxiv papers (21 billion tokens). For reference, this is almost half as big as the total training data that went into GPT-3. This is a 10x boost in technical knowledge that would exist inside current LLMs. It looks like DeepForest is engaged in the effort to make molecular foundation models and I’m looking forward to seeing what they create. When somebody does create this Ultimate Science Foundation Model (brand name pending…), and it may have already been done, I expect there to be a surprisingly powerful knowledge graph baked in that could take people years to explore. It is essential that such a trained model become a public good.

Thesis 3: There will be a universal language of physical science work that does not speak directly to humans. Monolithic cloud labs alone may not be optimal deployment of automated biology in the future. Projects like PyHamilton demonstrate growing open source communities for benchtop automation, and the SayCan collaboration by Google and Everyday Robots is a reminder of how multifunctional robots are steadily progressing (as well as ultralight indoor drones). As the cost curve goes down and the natural-language programmability goes up, there may be an intersection at which it is easier to convert an existing lab environment/protocol into an automated one, rather than to outsource work to a physically separate facility. Or, there may be a steady-state solution that some tasks are optimal for large automated warehouses and others are optimized for more distributed, edge labs. If there is any future of multiple robotic work providers, then interoperability will become a bottleneck, which will motivate a universal formalization of life science work.

Personally, I’m most excited to see how AI can help us navigate enormous wells of scientific knowledge (eg, our geochemical negative emissions technologies review) to propose and execute experiments that expand the frontiers of planetary-scale biotechnology. I think it’s important to get high-impact problems into the discussion that may not obvious short-term payoffs: medical applications are well-defined and financially mature4, whereas frontiers of energy, food and biological resilience are critical but under-resourced relative to impact. New AI abilities might change the calculations of costs/risks to make new spaces suddenly tractable: for example, ML protein engineering tools today are powerful but multi-protein complexes (eg, the exquisite nitrogen splitting anvil) have been out of reach. The next generation of computational tools, if properly integrated into experimental workflows, could help us operate at the horizons that Stephen Hawking saw in 2000:

I think the next century will be the century of complexity. We have already discovered the basic laws that govern matter and understand all the normal situations. We don't know how the laws fit together, and what happens under extreme conditions. But I expect we will find a complete unified theory sometime this century. That will be the end of basic theory. But there is no limit to the complexity that we can build using those basic laws. This complexity can be biological or electronic.
—Stephen Hawking in an interview, San Jose Mercury News, January 23, 2000.

To conclude on an emotional note: this is a phase change moment in time, the future is uncertain and only getting harder to predict. This is uncomfortable for many. It doesn’t help that the vast majority of our SciFi universes skew deeply negative (shoutout to the solarpunks and those trying to build positive outlooks), so let me leave you with the mental image of Glenn Gould back at his piano. He is older now, in the small room he chose over the palatial concert hall, in an environment he meticulously crafted to the degree of individual baffle positions5, surrounded by the best recording engineers in the business who had passed his quizzes. Glenn is sitting exactly 14 inches off the ground on the only chair he ever played on, hunched over a piano that is on wooden blocks to get his precise desired height. He is in his moment of flow born of decades of mastery dedicated to getting the perfect sound of Bach out from the piano and stored onto tape. And in this moment, Glenn Gould is singing his heart out.

Acknowledgements:
There are many amazing people I’ve discussed these ideas with over the past two weeks, and I hope you know I’m very appreciative of our conversations. Specific thanks to Matt B, Jacob A, Tom K, Erika D, Neiman M, Jose L, Paul R, Jake F, James K, and Eitan M.

Consider two large problems that share a non-obvious link:

1. 16% of human deaths globally are directly attributed to pollution today and 12% of radiative forcing will be due to industrial bi-products by 2050. Yet, despite these chemicals being tied to some of the biggest disease burdens of our era, there is relatively little direct effort to cure or reverse the diseases linked to these chemical injuries. 

2. As long as it is maximally profitable for a company to generate toxic externalities, the company is mandated to continue that activity. CEOs are personally liable to operate in shareholders’ interest, whereas the corporations are liable for environmental impacts.

If these two points seem unrelated, it’s because there is currently an Accountability Gap between the cause and effect of industrial byproducts. What if we lived in a world of perfect molecular information in which every molecule is accounted for and every downstream cascade is perfectly observable?

New technologies are needed to bridge the Accountability Gap: Solving the challenges of discovery and resilience could form one of the biggest fulcrums yet to drive global economies toward cleaner, more sustainable practices.

Intro: Shaping the problem

Human health is an under-appreciated lever for planetary health. When people study pollution now, generally via government research grants, they tend to study it from an epidemiological perspective. These grants become correlational studies with the assumed end goal of informing new policy. Frankly, I think this is slow, indirect, and worst, only going to decrease in efficacy as industrial technology outpaces regulatory infrastructure. We need to tap the power of the billion-dollar biotech machinery to go molecules-up on large-scale mechanistic studies to invent interventions to directly address the presence and effects of pollution. Success could be both profitable and impactful.

In the recent boom of ClimateTech, there has been an intense focus on building innovation engines to mitigate the three big sources of radiative forcing: carbon dioxide, methane and nitrous oxide. The technical challenges of management of these three gasses have increasingly actionable surface area and, specifically around CO2, rapid mobilization and innovation in funding has created hundreds of companies within just a few years. This nascent ecosystem has made huge progress (see my essays on gaps in early stage climateTech funding and a first rough draft on climateTech benchmarks), and I worry that blindspots are forming. Take the graph below as an example: It explains why CO2, CH4 and N20 are rightfully prioritized, but I also see 17+ gasses that make considerable contributions. What this graph does not show is that, on a per molecule basis, some of these “other”1 chemicals can have 100-10000x the warming potency of CO2. Importantly, this “other” category is increasing in magnitude and poorly understood in other biosphere impacts. So, generalizing beyond greenhouse effects, what about all the other industrial byproducts that haven’t yet received the same mass mobilization effort as CO2, CH4 and N2O?

The independence from policy and regulation made Negative Emissions Technologies (NETs) a perfect nucleation point for builders. Without being tied to the slow Sisyphean rocks of bureaucracy and population-scale decision making, NETs innovators have enjoyed a freedom to step in and chase scalable approaches to physical limits (e.g., toward the <$100 ton CO2 capture goal). Because of this freedom and the accessible technical challenges, carbon capture is a gateway drug2 for technologists to search for a match between their abilities and the big, meaningful problem of “Climate”3. Now, with the critical mass of funding support and growing community of activated practitioners in ClimateTech, we can explore other challenges that, in contrast to NETs, are intentionally intertwined with existing systems4. This means now is an opportune time take more technical risk, providing the contours of the chosen problem can fit existing gears of industry.

A prime first example of one such highly interconnected problem is the negative externalities of small molecule byproducts of industry, AKA pollution.

When I talk about pollution and biotech, I use the clickbait-ish phrase “Pollution is a Pharma Problem.” I’m being intentionally loose with the word “Pharma” to be broadly inclusive of the new reaches of industrial medical bioengineering5. More importantly, I’m showing respect for the enormous levers of biotech innovation around human health. My optimism is that we can discover ways to steer these billion-dollar biotech levers toward impact on planetary health.

Four interconnected subindustries

On its surface, the challenge of managing small molecule waste6 seems impossibly hard: tiny molecules in unknown states of degradation interacting in unknown mixtures with unknown biological effects in a variable timeframes in unmeasurable situations. Which slices of these dimensions are productive to access, what technologies are we lacking and how much data do we need?

The high level challenge for the biosphere is that industries will innovate and deploy novel chemicals faster than governments can discover and regulate them.  Especially as we’re in the early stages of a boom era of computationally generated molecules, and in mid-late stages of regulatory incapacitation (eg, “inert” ingredients don’t need to be declared except in California), I only expect the problems caused by toxicants to get worse. Even in the most optimistic futures, we must be prepared for the biosphere’s longterm cohabitation with anthropogenic chemicals, which will mean developing methods for resilience. For example, we all have the teflon precursor “forever chemicals” in our body now:

Even though there are active government efforts, and important new policy proposals, and mature efforts for bacterial/viral diseases, it is essential to also discover free market solutions to small molecule externalities. Due to significant technological advances in biology over the past decade, the time is right to develop industries around this challenge.

I see four industrial opportunities that can be partitioned into highly synergistic verticals and together represent a TAM in the hundreds of billions:

• Forensics: How might we significantly increase the scale, speed and generalizability of small molecule detection? Possible markets might be legal discovery and expertise (e.g., the $10.9BB Roundup lawsuit) and consumer exposure tracking. Projected profitability is low, impact to other businesses are high. 

• Remediation: How do we remove the toxin once we’ve identified the harms and the molecules? Total TAM is currently ~$100 Billion with CAGR of 9.8% (source). Environmental remediation was specifically identified in the Biden “Build Back Better” bill, with $9BB just for lead in water supply alone (link). 

• Neutral Alternatives: How can we accelerate the development and validation of bio-neutral alternative chemicals? Can a modern forensics pipeline that detects bad molecules also be used to develop cleaner replacement molecules? If so, are there IP licensing opportunities or other viable businesses?

• Therapeutics and prophylactics: Can we build on the biotech playbook of rare, monogenic diseases to create a medical industry around mono-toxigenic diseases? The monogenic disease playbook has been played out for dozens of billion-dollar biotech companies, what is missing to bring environmental toxicology into this world? Similarly, can we re-purpose the pipelines and tools that the pharmaceutical industry has developed?

When laying out these four industries, one can imagine many possible connections between them. For each industry, I will ask some questions that could plausibly be answered by biotech today. Perhaps in a later piece I’ll articulate that gaps that explain why so few of these challenges are currently attempted by industry (let me know if you’re interested). But first, I want to give a brief sampling of the issues out there today to motivate the incredible efforts needed in the future.

A brief history of gross chemicals and human health

In 2012 I was driving on highway 101 listening to the radio and heard the writer Florence Williams as she went on NPR to explain her new book “Breasts, a Natural and Unnatural History”. I was, and still am, captivated by her story. She wanted to understand why girls as young as 9 were prematurely developing breasts, so she and her daughter sought to rid themselves of all exogenous chemicals. Even with measures such as wearing old clothes, only bathing with Dr. Bronner’s, and avoiding plastic food containers, there were chemicals in their blood that could not be removed. Furthermore, she references studies on how the composition of breast tissue makes it a concentration site for toxins. She talks about many chemicals that are now familiar to us such as Bisphenol A (BPA) and PFAS. The lingering question is: what other chemicals are in us now and what effects are they having?

In a curiosity similar to Florence Williams in 2010, Professor Edward Kolodziej in Washington in 2018 was puzzled why the Coho salmon died every time it rained. Coho Salmon are keystone species, meaning they are an essential layer of the food pyramid, risking collapse of the forest and marine ecosystems if the Coho are irretrievably lost. After two years a heroic effort of analytical chemistry (for any mass spec nerds out there, six fractionation steps!), they discovered that a tire preservative was responsible for killing juvenile fish. Even more difficult, it wasn’t even a known chemical that had been registered with the EPA that caused the damage. The tire preservative 6PPD degrades to 6PPD-quinone, which is highly toxic for just the juvenile fish. The study was published in Science, caused a big splash, but two years later, 6PPD is still irreplaceable because there are no known alternatives.

Most people think that we know what chemicals are toxic, and all we have to do is control the amount of those chemicals to make sure water quality is fine," said co-senior author Edward Kolodziej of UW. "But, in fact, animals are exposed to this giant chemical soup and we don't know what many of the chemicals in it even are." (source: NSF)

Forty-seven percent of the 86,405 chemicals registered with the United States Toxic Substances Control Act (TSCA) inventory as of June 2020 are actively manufactured, processed, or imported. Furthermore, according to the non-profit Environmental Working Group, the federal standards for safe drinking water have not been updated for 20 years (for context, that was when Snake was the only phone app). And, even if we could assume that all existing industrial chemicals were registered, as we saw in the CoHo salmon case, chemicals degrade in the natural environment, creating downstream unregulated toxic molecules to unexpected corners of the biosphere.

Here is a quick list of some the best known and studied chemicals in our environment:

• Antibiotics in agriculture: The overuse of antibiotics for both industrial terrestrial farming and aquaculture has produced measurable amounts of antibiotics in critical waterways. Beyond antibiotic-resistant pathogenic bacteria, these chemicals are associated with cancer, bone marrow disease and immune dysfunction (see: the EU’s maximum residue limits and the health effects listed in the chemicals associated with fish farming ). 

• PFAS: The PFAS family is enormous (well known from Teflon production), containing over 5,000 different compounds with well-documented health efforts (see: the PFAS-Tox database). 97+% of Americans have PFAS in their blood and now one science team is claiming all rainwater in the US is undrinkable due to PFAS. 

• Fracking chemicals: One 2017 paper from Yale Med studied the carcinogenicity of 1177 water pollutants and 143 air pollutants: 55 of which were known carcinogens, 20 of which had leukemia/lymphoma risk.

• Phthalates: the “plasticizer” chemicals found in beauty products, foods, mechanical oils, etc.   

• Organic solvents: TCE is a chemical known to directly impact the nervous system, yet is known more for its carcinogenic properties when it contaminates drinking water (source).  

• Concentrated metals: Heavy metals are essential for industry and unfortunately have the ability to spread, only to be re-concentrated up the biological food chain or pristine forest. Mercury can be spread over the air and rain down to concentrate in unexpected places like pristine forest. Lead is the most well known of all: “The State of Michigan considers all children in the City of Detroit to be at-risk [of lead poisoning].”

• Insecticides. Organochlorine, for example, has been the most studied and, while banned in developing countries, it continues to increase in usage around the world . There have been a few exciting new biotech companies founded in the past few years beginning to address this space (eg, Greenlight Bio)

• Phthalates and other EDCs in consumer goods: Phthalates are the “plasticizer” chemicals found in beauty products, foods, mechanical oils, etc. and one of the big Endocrine Disrupting Chemicals in the world today. The CDC says detectable phthalate metabolites are  “widespread throughout the general population”. “EDCs can impact the endocrine system and subsequently impair the development and fertility of non-human animals and humans” and includes “ disinfection byproducts, fluorinated compounds, bisphenol A, phthalates, pesticides, and estrogens” (Gonsioroski, Mourikes, and Flaws 2020). An unexpected canary in the coal mine could be the unfortunate polar bears, which have been impacted by the hormonal waste of humans that are transported by ocean currents, creating a high rate of hermaphroditism (Nature blog, 2003). 

To give a sense of how little is understood about the health effects of small molecules, consider that while herbicide glyphosate does not directly cause obesity in rats, nor their offspring, it does cause obesity in their grandchildren and great-grandchildren (who had no direct exposure). This is very strange and shows us how little we actually understand about biological systems. But by understanding the effect of harmful small molecules, it’s an optimistic possibility we could also learn to work with positive small molecules. For example, certain microbial metabolites in the gut are linked with human lifespans of over 100 years. The rapidly growing strengths of bioengineering will continue to broaden the scope of addressable problems and opportunities, it just needs the right focal points.

Fertility is one example of a humanity-scale critical problem with strong links to pollution exposures. We know that there is a net decrease in testosterone of about 50% over the past 40 years, including a 25% drop from just 2000-2016. Furthermore, studies of sex hormones have shown direct links to small molecule exposure. Phthalates, for example, have been linked to a 24-34% decrease in testosterone among young boys ages 6-12 . BPA is detectable in 92% of the US population and also associated with a significant decrease in testosterone. Malformation of penises in utero are also associated with endocrine disrupting chemicals. For an extensive overview of exposures and male reproductive health, refer to this 2018 review, ItStartsWithTheEgg.com or MillionMarker.com.

This list of chemicals and diseases is of course non-exhaustive, and I think many of us have been saturated by these headlines. So where is our sense of technological optimism and entrepreneurial energy ?

To spur conversation, I’m going to ask technical questions inside each of the four industries and rank them by how technically far away we are from them today. “Easy” means we are technically able to do it today, “hard” means we would need a concerted effort to accomplish an answer. And recall the opening two thoughts to this essay: 16% of global deaths are caused by pollution and there is a wide gap of accountability, so our goals here are to imagine futures in which the economics of pollution could be meaningfully changed.

Industry 1/4: Forensics

How might we massively increase the scale, speed and generalizability of small molecule detection?

Easy: “Are specific Chemicals A, B and C in my water? Is A, B or C in me now?” (examples: MillionMarker for consumer testing, MatterWorks for AI + Mass Spec).

Medium: “If direct detection of the molecules is either prohibitively expensive, too time sensitive or insufficiently robust, what are associated biomarkers like RNA or stable proteins that are easier to probe? Can it tell me if I was recently exposed?”
(examples: serum markers for PFAS exposure, LifeExtension.com sells a $150 urine test for 8 exposure metabolites)

Hard: “What are all the unexpected molecules in my McDonalds cheeseburger?”
(examples: one example of potentially rapid progress happening in untargeted mass spec, but the scope of the challenge is enormous)

Bonus: Can we create molecular fingerprints to pinpoint the original source of the pollutants? If fingerprints are combined with ways to rapidly discover causation of disease from a pollutant, do we create a fundamentally new legal path to holding polluters accountable?
(examples: tracing PFAS soil pollutants to the source, Phylagen’s supply chain monitoring, the $11B RoundUp lawsuit came down to proving causality, and the incredible story of Robert Bilott which became the 2019 movie Dark Waters)

Industry 2/4: Remediation

How do we remove the toxin once we’ve identified the harms and the molecules?

Easy: How do we remove chemical X from a point source?
(examples: removing the irreplaceable 6PPD-quinone from urban runoff or glyphosate from farm runoff. Epoc Enviro has an elegant PFAS filtration method, now licensed by Allonia)

Medium: Can we make a product that removes chemical X from a natural environment, rather than selling a CapEx intensive cost-plus service?
(examples: plants are metal hyperaccumulators, Gerben Stouten’s Platicacumulens microbe eats trash)

Hard: How do we get rid of PFAS from environments or organisms?
(examples: according to the Secretary of the US Navy, there are no known methods of getting PFAS out of the human body and across the food web some PFAS chemicals are biodilutive while other are bioamplified. )

Industry 3/4: Neutral Alternatives

How we declare or invent a molecule to be completely bioorthogonal?

Easy: Does known molecule X have any interaction with Druggable Protein P7?
(example: organophosphate pesticides bind to acetylcholinesterase )

Medium: Does candidate molecule X have any interaction with any druggable protein or other mechanisms of bio-concentration?
(example: the Warner Babcock Institute for Green Chemistry or the SaferMade portfolio )

Hard: Does candidate molecule X, and any degradation product of X, have any interactions with the known biosphere?
(example: virtual screening has been longstanding effort in the pharma industry, VirtualFlow as one example, but instead of screening billions of candidate drugs onto one protein, it would be one molecule and its derivatives against the addressable proteome)

Bonus: Are there first-principle criteria by which future industrial molecules could be designed in order to create minimal biological impacts?

Industry 4/4: Therapeutics

How do we heal people that have unavoidable exposures to bad chemicals?

Easy: How can I minimize damage from a known molecule with a known toxicity mechanism that causes a known disease8?
(example, bioclearance is well-studied by pharma companies and clinical pharmacology labs)

Medium: How can I rapidly and repeatably discover the molecular-scale pathogenesis of diseases caused by a mixture of one or more toxicants?
(example: exploring Parkinson’s Pathogenesis in Drosophila)

Hard: How can we build the field of toxicogenic diseases9 in the same way that rare monogenic diseases and subtyped cancers became the foundation of the the $100Billion+ precision medicine industry?
(example: gold standard datasets and tools such as the offerings from the Broad Institute or OpenTargets have been an essential resource)

Conclusion (for now)

There is obviously much more to say than the kernels planted here. The point is that there is tremendous much room for impact: beyond “just” helping us clean up our current environment by solving today’s accountability gap, building the biotechnologies to increase our resilience is a subtle but essential foundation to optimistic futures such as space travel.

There is a community we can build around the current efforts on toxicology and environmental remediation. There are practitioners who will read this and immediately see answers to these industry questions. There are knowledgable leaders who may have insights to the magnitude of gaps ahead in order to build the translational field of toxicogenic diseases. There are also clinicians that have patient populations with currently unaddressable illnesses. I’d love to talk with all of you and see new networks form around this opportunity space: if anybody wants to go deeper on any of these directions, comment below or reach out to me via Twitter with request for a follow-up piece or discussion.

And if we do nothing? I contend that government efforts are essential but alone will not be able to keep up with the rapid pace of chemical innovation from industry. The status quo path is that we will co-exist with chronic perturbations that will certainly have the most impact on today’s youth (see the image below and this EA Cause Exploration Prize). If we do nothing, the massive boom of new biotechnologies goes under-utilized on today’s narrow scope of industrially addressable problems. This could be one of the biggest and most direct opportunities for the biotech/pharma industry to achieve both positive environmental and social justice impact, and I aim to see this potential fulfilled.

Acknowledgements

Huge thanks and appreciation to all the amazing people who have discussed this with me over the past year Jenna, Josh, Ethan, Tony, Scott, Georgia, Loren, Sam, Sam, Erika, Adam, Anastasia, Milan, Tatyana, David, Daniel, Bob, James, John, Alexander, Henry, Carly, Brett, Elliot, Willy, Jenny, and many more. Special thanks also to Hanny for playing the role of friendly editor on this piece and finding many opportunities for improvement.

This time last week I was jotting down notes in a hotel after an amazing two days at the Funding the Commons conference organized by the Protocol Labs. There were many great experiences and new connections made, but two highlights stand out: It was our public debut of our non-profit Homeworld (“Hello, (home)World!”), and the first time I began to see potential in this wild idea of Impact Certificates.

First, If you are interested about Homeworld, you can see my talk on behalf of our community, and screenshots of highlight slides, in Vincent’s tweet below. While most people at FTC22 were talking about platforms, I was talking about content that might one day utilize such platforms. This dynamic made for many great hallway discussions, including the idea of Impact Certificates.

The idea of Impact Certificates goes back to 2014 in the Effective Altruism forums. Going through those old forums now, the ideas sound interesting but wouldn't have stood out to me (an outsider to the EA community) as particularly groundbreaking. But Juan Benet’s talk presented a very cool viewpoint which stuck with me as compelling and worth further exploration, especially from the perspective of trying to fund early stage biotech projects for the public good of climate-positive tech development. Again, Vincent’s tweet is a good entry point:

Juan’s talk explores why the venture capital (VC) model has worked so well well1. One reason is that the VCs who make the actual investment into a company tend to be close to some aspect (technical, social, etc.) of the projects that they fund. The VCs exchange stock certificates for cash with the startups, and then the VCs pass the stock certificates onto their own investors who then have their own investors, etc.. It is non-obvious that more steps in a network is a good thing: why would all these intermediate steps between the company and the original sources of the capital make sense?

There is a topological argument of why this VC structure is powerful. Juan points out that this looks a lot like the composable learning units inside a deep neural network. Deep nets excel at learning complex solution surfaces at multiple scales (ie, broad trends vs tiny details), which Juan further illustrates in the slide below.

Composability enables multi-scale learning and execution, which is what enables financial behemoths like the state pension funds to profit from young startups. Let’s use CalPERS as an illustrative example: their $470BB fund gets distributed to different managers with diverse theses, some of those managers invest into VC funds for their diverse sub-industry specializations, and finally each VC finds+supports the best startups for their given zone of competence. Profits are ultimately returned to CalPERS via stock certificates, irrespective of how many intermediate layers. There is indeed a commission paid each node hop between startup and primary funder, but the costs are a tradeoff for distributed sector expertise. For playful contrast, imagine a CalPERS employee going through Hacker News forums in an attempt to deploy $1Billion into startups: that money would get vaporized with no meaningful impact. Hence, the composable funding structure, enabled by the (eco)system of stock certificates, facilities large amounts of capital to be matched with expertise that finds good allocation opportunities. It is certain that, regardless how indirectly, money from enormous funds have been responsible for expanding the frontiers of science and technology.

The composable funding structure then sounds like a good idea, so the natural followup is to ask why it is not deployed in every funding environment. In the absence of profit incentives or financial infrastructure, the stock certificate (eco)system does not apply and composability feature breaks down. For example, if you were a Gates Foundation grantmaker with the mandate of saving lives via malaria medication (a very different mandate than the profit mandate at CalPERS etc.), you would likely be directly funding the organizations who distribute the malaria medication themselves2. This is because you would need to validate that you’re funding is actually reaching its intended impact. So, if you had $1BB to spend in a year, you would just be talking mostly with the entities most credibly ready to spend it immediately in an auditable way, which then may force too much money into the limited pool of candidate entities to fight malaria. Should you instead try to go through arbitrary numbers of middlemen in a foreign country in a vain attempt to replicate the like the CalPERS example above, you lose oversight ability and would feel at high risk of being lied to. How would you have any second-hand trust in an organization that you’d never met or validated yourself? How would you report your performance to your boss?

The bold idea to explore is what if impact certificates could be the non-profit version of stock certificates? What would it take for a certificate to enable the same efficiency of profitable investments into the public good investments? Juan suggests three features to start:

• Feedback: The performance of the underlying asset is the same for all holders of the certificate. Information has to propagate through the network in order for the ecosystem to improve.

• Transferability: The certificate is able to be passed throughout the funding ecosystem

• Fractionalization: The certificates can be split up for compensation and reselling.

Before going into my scientist’s perspective questions on what Impact Certificates might need to work in our domain of biotech, it is worth unpacking two other talks from Funding the Commons that articulate the aspirations of Impact Certificates.

David Dalrymple pointed out in his FTC22 talk on “hypercerts” that there is a temporal component of impact certificates. If a public good task has been completed, then there is no reason for a funder to pay for it: As on explantation for why this is the case, consider that there is no additionality in funding something that has already been done. As David says, what you want is wormhole for the funder to know the contributor is going to do that good and so fund it. Or, easier than a time machine, what if there was a class of investor looking to catalyze public goods at a lower or equal rate to what the ultimate benefactor would be. This might be a step towards that composability idea: the primary philanthropist could buy the right to say they enabled that public good to be created. Beyond the temporal dimension, Impact Certificates may also help with breadth of discovering good uses of funding.

Karola Kirsanow articulated the frontier of graph-based roadmapping in her FTC22 talk, which seems to have a natural future integration with Impact Certificates. Roadmaps are of particular interest in neuroscience and climateTech, where most of the problems are enormous and complex. For examples on roadmapping, refer to TheClimateMap or this recent perspective paper on mineralization for carbon capture with the biotech section led led by Paul Reginato. So, in Karola’s talk, she presents the idea that if you have a knowledge graph, and a data model for a scientific process within a node in that graph3, the marketplace for funding could be built directly into the frontiers that graph. That’s a really big idea.

So here’s the hard part: details. What the heck does this actually look like in practice? For this post, I only want to sketch some thoughts of what Impact Certificates might look like in the context of funding biotech projects for the public good of the environment.

Let’s think through the simplest example. Assume you have a Project C, funded directly by an Edge Funder4, which distributes funds from Primary Funder P. To make this example tangible, let’s assume C is a bioengineering research effort toward a scalable carbon capture approach. For Project C, the core research risk is harnessing a previously difficult-to-engineer organism, and if the risk is overcome, there is a highly fundable larger project (or company) to be built. The interactions look like this:

So, if I understand the vision of Impact Certificates, the perfect world scenario is Project Team C does amazing work and the corresponding impact certificate becomes something worth transferring. Project Team C’s success at de-risking their effort makes them eligible for other existing sources of capital, which means the funding behind Team C was additional. Other Funders are now interested in purchasing ownership of the Impact Certificate for Project Team’s C initial project from Edge Funder E. This purchase of the Impact Certificate effectively frees up Edge Funder E to now make more public goods investments in the future5.

Even in that perfect world scenario, it raises a lot of specific questions. Below are my top 6 questions with some possible answers.

1. When is the impact certificate granted? A stock certificate would be upon the transaction of cash from the VC to the startup, so let’s assume the default answer is an impact certificate would be created upon Project C’s funding from the Edge Funder.

2. Who validates the things that Project C or the other owners of the Impact Certificate claim happened? The simplest answer is there is no automated validation at all, and any buyer protection is the responsibility of the people who transact in impact certificates. People who buy stock certificates are expected to do their own research but rely on basic conventions of good corporate behavior. So it may be possible that procedural validation of the Impact Certificate might be in everyone’s best interest: consider the damage of fraudulent pumping of the impact6 by any of the parties involved. Or, if we revisit the toy example above of Gates Foundation trying to deploy $1BB across multiple steps of organizations for malaria, some validator would be an essential piece of the Impact Certificate ecosystem.

3. Would it have the same feeling of philanthropic success for a Primary Funder to re-purchase the Impact Certificate from something they did not originally fund? This is a core assumption behind the idea of reselling Impact Certificates and could be validated by conversations and prototypes. Say Project C becomes a breakthrough in climateTech: Edge Funder E would be incentivized to sell their Impact Certificates in order to then fund future Projects F, G, H… But would it appropriately match the mandate of a major foundation like the MacArthur Foundation or the Simons Foundation to repurchase Project C’s Impact Certificate? In 2022, my guess is the answer is generally no, but it may be meet the needs of focused efforts. For example, Stripe has done great work in funding carbon capture efforts, perhaps they would repurchase the Impact Certificate for Project C with the intent of freeing up Edge Funder E to make their next bet.

4. What details need to be published in the Impact Certificate? If the impact certificate is indeed funded at inception (question 1), then the details of the project would basically be the grant application that acquired the funding. This is somewhat unsatisfying because the most interesting details will come from the development of the project, but that would require updating the Impact Certificate, which may be undesirable.

5. What if Project C becomes a company? Say in two years the team behind Project C makes a huge leap forward, how can that leap be attributed to the original impact certificate? While it sounds shiny to think “woah maybe an Impact Certificate could convert into a stock certificate!” that feels very difficult and very complicated. I think the answer has to be to stay as absolutely simple as possible, in which case there is no formal arrangement for Impact Certificates V1 to have anything to do with downstream IP or commercialization.

6. Who benefits from Impact Certificates? Funders. Primary Funders with a clearly articulated target portfolio of impact might like the ability to observationally de-risk projects before “retrocausally” funding them (David Dalrymple’s word). Edge Funders would love Impact Certificates because the new marketplace would allow Edge Funders to exist. And, maybe, the Project teams would also like the Impact Certificates as presumably they would hold on to some fraction of that ownership too.

Are there any precedents it Impact Certificates? At first glance, it feels roughly most similar to the Carbon Credit marketplace. The Carbon Credit marketplace can be a rough comparison: at its worst, Carbon Credits are murky, filled with scams and unintended backfires. However, failure in practice is not proof of flaws in the theory and successful carbon credit system could be one of the biggest levers we have in mitigating climate change. Failures aside, there is a reason huge amounts of money is flowing into the Carbon Credits, hopefully soon to be followed by substantial innovations in supply, demand and market security (eg, verification).

So I leave the reader with open questions about Impact Certificates. Would you accept funding that came with an Impact Certificate? Would it have any effect on your feelings around the funding and your publishing strategy? Are there exciting frontiers or fearful pitfalls you see in this idea?

1

The top VC funds have returned famously

epic IRRs, but keep in mind that the power law of returns applies to VC firms just as much as startups. The median return of a VC fund is probably closer to breakeven (see this 2012 analysis). What matters in this context is that the best VCs have built and shaped industries that became significant fractions of the modern economy.

This is a work in progress that I’m sharing broadly in hopes of generating feedback, ideas and community discussion. All thoughts and input are appreciated, especially as depth beyond my own expertise is needed to craft the most impactful benchmarks: the easiest way to reach me currently is via Twitter. Success in this effort is help onboard new people into the field of climateTech, with an emphasis on the bioengineering perspective. Together, I hope we can create shared targets that drive innovation, and support new funding efforts that can move at the speed of experimentation.

Goal

As more and more people look to get involved in climate technology development, it is essential that we craft entry points and shared community goals. I’ve argued previously that all hyper-productive communities share common attributes, and here I want to explore the important of well-articulated shared goals. Many people, including myself, spend months and years just trying to find the right problem to work on. One lightweight solution to a “where do I start?” problem is to establish public targets for people to nucleate their early stage ideation. Success in this approach would be a set of quantitative benchmarks that aspiring entrepreneurs and scientists can aim to beat. 

This is a work in progress: We intentionally put “V0.1” into the title to welcome community-wide input and conversation. There are creative tensions in this work, which we discuss at the closing section.

By establishing target goals, we can broaden the base of technologists working on this problem and direct funding to support creative, novel efforts.  

Criteria for a good benchmark

Good benchmarks both activate a community and illuminate an individual’s first steps. We used the following metrics to assess quality throughout the process of assembling and editing these benchmarks. 

• Tangible and actionable, even for a non-expert. Is it obvious what this number means and why is it important? If not yet we deem it is important, what information can we provide to most quickly get them up to speed?

• Measurable on a minimal unit basis. What is the smallest experiment that could be run based on this benchmark? In the case of a crop plant like corn, “~50% of its nitrogen is synthetically fixed” (Bloch et al 2020) focuses on a single plant, but might be too granular depending on the state of the art of the experimental assay. The next step up in unit size might be to say “75lbs of N fertilizer per acre” (source: EPA 2019) and is perhaps the right unit to work with, whereas the largest unit, e.g. “total synthetically fixed nitrogen usage in the USA is 20 million tons,” may be interesting but unactionable (source). 

• Specific and easily replicable. Let’s say, for example, we eventually make a benchmark for nitrogen fixation in organisms. If the best nitrogen fixing organism’s N2 reduction reactions are measured at ~10^4 /s/cell , a good benchmark would reference the paper, including the organism and assay. If a benchmark is not easily repeated in a lab, for example $300/ton of CO2 captured at atmospheric conditions, reference the best available publication.

• Maximally future-proof. Benchmarks are naturally going to have a shorter horizon than roadmapping towards theoretical ideals, but good benchmarks should be maximally agnostic to current technology and with assumptions listed as explicitly as possible. 

Scope of the benchmarking project

We pragmatically accept that not all benchmarks are equal in impact and mindfully avoid making prescriptive hierarchies. To properly determine which problems are more important than others is the domain of careful roadmapping with diverse stakeholders. 

Consider, for example, the benchmark of 1 pound of synthetically fixed nitrogen via Haber Bosch (HB) per bushel of corn. Replacing HB-based ammonia with lower CO2 emission processes of nitrogen fixation would certainly be an important lever, yet one might argue that decreasing demand for synthetically fixed nitrogen is the “more important” benchmark. Such debates around hierarchies of impact are important but can quickly get mired in assumptions, proprietary information and projections: this is important but becomes a much larger project. As such, our principle here is that we need a portfolio of approaches for all the problems which we quantify in these benchmarks. 

What are examples of successful benchmarks?

Clear goals for a field are both rallying points for community and entry points for outsiders who may bring valuable expertise. One could argue that the very public and tangible number of $/ton in the carbon dioxide removal space has helped many potential practitioners start their brainstorming. What are other cases in which clear benchmarks helped propel a field?

• Protein structure prediction. For years, CASP was an annual event that challenged the structural biology and computer science community to predict the structure of a protein based on its sequence alone. The core quantitative benchmark was the Global Distance Test, with several variants. This challenge built around a benchmark attracted bigger and better efforts until in 2018, Google’s DeepMind entered the CASP13 competition and outperformed every other approach to such a degree that Mohammed AlQuraishi (a leader in the space) was compelled to blog “What just happened?” Since then, biotechnology has been permanently changed by access to estimates of previously unknown protein structures. 

• Autonomous Vehicles. Developing software that could autonomously drive vehicles in natural situations seemed like a long shot in the early 2000s when the DARPA Grand Challenge was announced. The milestone was to complete a 100+ mile course, but the benchmark was the longest distance the best performer of the previous year was able to navigate before getting stuck. In 2004 the best team accomplished 7.32 miles of a 150 mile course. In 2005, five vehicles successfully navigated the complete 132 mile course.

• Machine Learning Challenges. The Netflix Prize was the first high-profile public benchmark challenge in modern computer science. Announced in 2006 with the then-unbelievable prize of $1MM cash, the challenge was to beat a benchmark algorithmic performance by at least 10%. Over 20,000 teams got involved, and the target performance was accomplished in 2009. Also in 2009, Fei-Fei Li and her team developed ImageNet, a landmark database and annual 2010-2017 competition that became the foundation of the Deep Learning revolution. 

• Athletics. Consider the 4-minute mile in 1954, the 2-hour marathon in 2019, 1000lb deadlift in 2006. 

• Genome reading and writing. Cost per sequencing was highly tangible to the biotech community and as such, incentivized much focus and much funding. With the advent of CRISPR, there was a race in the protein engineering field to optimize the Cas9 enzyme to access as much of the genome as possible (cite Pranam’s Cas9). And in future genomic engineering challenges, such as writing whole chromosomes, Ostrov et al 2019 articulates the clear milestones of the genomic writing community.

Benchmarks

The V1 benchmarks will have four primary topics: Greenhouse Gas Removal, Materials, Agriculture and Energy.

Greenhouse Gas Removal

CO2 capture cost at atmospheric concentrations ($/ton) 

Why this matters: Gigaton-scale carbon capture is needed urgently and marginal cost is essential. According to the CDRPrimer.org Section 1-1, the hard-to-avoid emissions for human society in 2022 is on the scale of 1.5-3 gigatons annually, meaning even in optimistic policy and behavior change scenarios humanity will need billion-ton-scale carbon dioxide removal (CDR). <$100/ton has been a de facto target across the CDR community (refer to this entrepreneur’s guide to DAC from 2019). As it is unlikely that a single CDR approach can reliably reach and sustain gigaton carbon capture worldwide, we will need a portfolio of good approaches with different strengths. As one example of recent innovations, Professor Jennifer Wilcox’s oxide cycling (McQueen 2020) indicated a $46/ton capture might be possible, which is now being brought to market via the company Heirloom. Note that CDR approaches must be subject to the developing field of Measurement Reporting and Verification (MRV) standards established by the CDR community. Roughly, a gigaton at today’s rates might cost $1 Trillion, which would be 1.1%  of global GDP.

Today: ~$300-$600/ton CO2 via Direct Air Capture (DAC) at scale is the unpublished consensus of the DAC community. 

Target: <$100/ton CO2. This proof of concept was established by Keith et al 2018 but not yet commercially demonstrated as of May 2022. <$100/ton would make a DAC company profitable at least in California where the carbon credit is $200/ton (World Resources Institute 2022)1.

CO2 capture energy expenditure at atmospheric concentrations (GJ/ton)

Why this matters: When designing carbon capture technologies, there may be scenarios in which energetic efficiency could be a primary variable to optimize. For example, in privileged environments with cheap or free power, energy-efficient designs could shine.  The Climate Technology Primer series covered the theoretical energetic bounds of capturing the CO2 gas from the air  (see Section 2.1), which references a target efficiency of 30% on top of theoretical minima of 0.7GJ/ton. For example, when Sahag Voskian and Professor Alan Hatton at MIT first published their electroswing technology in 2019, which later became the company Verdox, they achieved a head-turning per 1GJ/ton but in conditions of 10-100x atmospheric concentrations2. To put this all in context, the global power usage is 83 Exajoules, so without further innovation, capturing a gigaton at 10 GJ/ton would require an infeasible 12% of the worlds power budget.  

Today: 5-10 GJ/ton CO2, as reported by the International Energy Agency in November 2021 

Target: 2GJ/ton CO2. Source: MacKay 2008 The theoretical minimum energy for capture would be 0.7GJ/ton at 100% efficiency, and in Without Hot Air MacKay estimated 30% efficiency as a reasonable best target. Storage energy cost can vary significantly by instantiation.

Silicate weathering rates (mol CO2 / m2 s) 

Why this matters:  The silicate-carbonate cycle, referred to as rock weathering in the CDR community, is a natural gigaton-scale carbon flux per year. Silicates reacting with CO2 to form carbonates is perhaps the biggest natural lever the earth has for gigaton-scale capture and storage (Ciais et al 2013). While there are some big ideas, there is much innovation necessary and a  perspective by Keleman et al 2020 is a good introduction. Briefly, the primary weathering reactions are silicate dissolution and carbonate precipitation, and the optimal conditions (eg, pH) for the two reactions differ. There have been some efforts to accelerate the weathering process, referred to as Enhanced Rock Weathering. Examples include such as distributing ground basalt across agricultural fields (Beerling 2020) or Project Vesta’s strategy of deploying olivine onto coastlines for waves to mechanically break up the rocks. There have been initial experiments using microbial co-culture (McCutcheon 2021) , siderophores (Torres et al 2019) or carbonic anhydrase in a reactor context showing modest rate increases that are not yet practical (eg Power et al 2016), but there is not yet a leading paradigm toward order-of-magnitude increases in weathering rates. Choice of material is important (see Renforth 2019 for a review on alkaline materials)., but acceleration is still needed.  In nature, there is evidence that fungi, trees and lichens can accelerate rock weathering, but it has yet to be harnessed at scale.

Today: Natural weathering rates 10^(−10.53) and 10^(−9.86)  mol/ m^2 s^1 for basalt and dunite, respectively (Strefler et al 2018). Basalt is chosen because of its high availability but generally low reactivity, whereas dunite has the one of the highest CO2 sequestration potentials.

Target: No consensus target exists yet, and the target can be considered in two forms: prioritizing speed or scalability for a process. In small volumes at high temperature and low pH, you could trivially increase the weathering rate by ~100x, but this is not a viable solution in practice. Hence, steps toward a >100x acceleration for a scalable process in closed reactors or 10x improvement over large, open volumes would be transformative.

Materials

Energy to break rocks into <10 µm particles (kWh/ton)

Why this matters: Increasing the surface area of a given rock is critical to carbon capture and metal mining processes: finer material can react faster, be used in more versatile processing environments and the impact of negative feedbacks (such as passivated surfaces) may be minimized. For an excellent introduction, we direct the reader to ARPA-E Program Director Doug Wicks’ OPEN 2021 talk on Energy Efficient Routes to Comminution. 10 µm is selected as a candidate desired grind size based on the tradeoff of energy input vs speed of weather (Rinder and von Hagke 2021). However, under 10 µm particle size, 90+% of energy is lost as heat in the milling process: with existing technology, the comminution of the rocks are bottleneck to ERW CDR, both in operating and capital expenditure. Alternatively, there are natural mechanical weathering forces that are both biological (eg, plant roots cracking rocks) or abiotic (eg, water freezing cracks rocks). Additionally, there are many examples from nature of biology weakening or dissolving rocks, such the Red Alder Tree Perakis and Pett-Ridge 2019 or a common brown mushroom (Pinzari et al 2022). Rock comminution may be an obscure problem but could unlock scalable weathering-based carbon capture methods and lower-footprint mining operations, both of which are essential for a stable climate and an electrified industry. 

Today: Olivine is 180kWh/ton (Rinder and von Hagke 2021), can exceed 1000kWh/ton depending on the rock composition in mining (Jeswiet et al 2016) 

Target: Not defined. Roughly speaking: A proof of concept that demonstrates a path to significant improvement over the status quo would be of high interest. 

Carbon-Negative Performance Materials (CO2/ton)

Why this matters: Decarbonizing construction is an opportunity to simultaneously reduce emissions and to sequester captured CO2 in the process. Wood is both a reliable building material and long-term carbon store, and there has been some efforts to increase the performance characteristics to become comparable to aluminum alloys (Xiao et al 2021). Cross-laminated timber is used today in construction and has been studied for life cycle analyses (Anderson et al 2022), and one conclusion is that increases of usable biomass will be essential. Similarly, Arnold et al 2020 produced a detailed techno-economic analysis for carbon negative carbon fiber via algal production of the polyacrylonitrile feedstock: they indicate potential for carbon fiber to reach the gigaton of carbon storage, providing algal production advancements can be discovered. As steel is 8% of the global greenhouse gas emissions, any viable replacements or footprint reductions should be explored  (McKinsey 2020). At a similar scale, concrete is also 8% of the global GHG emissions, and carbon negative concrete has been a vibrant area of entrepreneurship (Nature editorial, 2021). More speculatively, there are many performance biomaterials from nature at various stages of study and deployment, such as spider silk, insect chitin or mollusc nacre. Inventing new concepts for carbon-rich, low-footprint materials could have a large carbon impact and may be best developed in collaboration with the construction companies. 

Today: Steel emits 1.85tons of CO2 for each ton of steel (McKinsey 2020). Carbon Fiber is <.01% of steel’s volume and 20x the price, and each ton of carbon fiber emits almost 30 tons of CO2 GHG equivalents  (source: Composites World 2019).

Target: Construction-ready material with a negative GHG footprint. 

Agriculture

Edible calorie production rate  (calories/acre/year) 

Why this matters: Scalable, distributable and robust food systems are a critical component of the technologies that humanity will need to live in the next 30 years. Focusing on calorie production density could help drive experimentation in cellular agriculture, indoor agriculture or increasing yields of existing plans. Focusing on calorie production rate could also drive innovation for stopgap solutions in dire situations of famine. As complex supply chains show fragility, and as changing weather patterns raise concern over traditional agriculture’s security, the time is right to consider new food production stacks. Organizations like ALLFED or New Harvest have been leading this new path, and more technology innovation is needed to enable decentralized production and resilience to unforeseen volatility such as drought and pests. One could start by modifying plants that we already know how to industrially process, or to start from organisms that have phenomenal growth rates. Consider the potential of fast growing cyanobacteria that divide 16 times in a day (up to 65,000x biomass increase), or that duckweed is thought to have grown so fast as to cause an ice age (The Azolla Event), that have not yet been deployed at climate-relevant scale.  It is possible that future paths for food production will be able to produce more food at a lower carbon footprint.

Today: Corn produces 15 million calories per acre per year (Washington Post, 2015)

Target: Can we beat corn on the acre scale anywhere on the planet?

Crop dependence on synthetically fixed nitrogen (pound synthetic N / bushel crop)

Why this matters:  Roughly 50+% of crop nitrogen comes from synthetically fixed nitrogen3. The denser the target crop yield (bushels/acre), the more the dependence on fertilizer (example). Synthetically fixed nitrogen has an enormous footprint: 1.2% of global CO2 is from Haber Bosch at a rate of 1.8tons CO2 for each ton of synthetically fixed nitrogen as ammonia. Furthermore, excess Nitrogen in soil has downstream effects, such as local toxic algal blooms and nitrous oxide release.  50% of the total anthropogenic flux of nitrous oxide, which is 300x more potent than CO2 and the number three source of radiative forcing on Earth, comes from agricultural fields and is strongly linked to fertilizer rate (Shcherbak 2014). In 2008 it was forecast that nitrogen-use efficiency could be increased by 50% (Erisman 2008), but some data indicates we are trending toward more, not less, fertilizer usage (FAO 2022 Report). There is significant room for innovation around reducing usage of synthetically fixed nitrogen, increasing naturally fixed nitrogen, preventing nitrous oxide formation, and engineering self-sufficient plant designs (Anas et al 2020 was a decent technical introduction in the context of sugarcane).  It also raises fundamental questions about plant biology: Can nitrogen fixation occur anywhere on a plant, why is it only the nodules in the roots that nitrogen fixing symbioses occurs? Could it be done in  leaf cells or in emulsion? 

Today: Corn is about 1 pound fertilizer per bushel

Target: Not yet defined, anything substantially better and scalable is important.

Energy efficiency of biomass production (Edible calories / Watt)

Why this matters: The incredible increases of crop yields in the 20th century, primarily via the proliferation of synthetically fixed nitrogen and heartier staple crops, have stagnated. For biomass-based production of food, chemicals and energy to realize its trillion-dollar potential, pathways to massively scalable biomass production will be needed. Phototrophic mechanisms are undergoing rapid innovations that may reach deployment: consider the TaCo and CETCH cycles invented by the Tobias Erb and his collaborators, the ability to screen Rubisco using EColi by Robbie Wilson, or in 2021 Cai et al published 7% photo efficiency into carbohydrates. It’s also plausible that hybrid electrotrophic paths might play a role in future biomass production. In 2016 the Nocera and Silver labs published The Bionic Leaf claiming a 10% CO2 reduction energy efficiency in small reactors. Finally, there may be a future of entirely electricity-driven biosynthesis, in contexts such as space travel (Martínez et al 2021 contends 10-21% energy-to-food efficiency), and it has been hypothesized that if oxygen sensitivity can be solved, yields could exceed 50% (Salimijazi et al 2020). This raises speculative bioengineering challenges: would it be possible –and beneficial– to do transplant metabolic pathways? Eg, to put the Shewenella direct electron transfer pathway into deployable yeast or fast growing vibrio natriegens.

Today: Modern crop plants are 1-2% efficient, C4 plants are 2.5% efficient (Zhu et al 2010)

Target: Not yet defined, anything substantially better and scalable is important.

Use of pesticides on corn and soybeans (pounds/acre) 

Why this matters: Protection of human and ecosystem health has been underappreciated, relative to its potential impact, by the climate technology community. Just as there has been a concerted effort to monitor, mitigate and sequester the unintended excesses of methane and carbon dioxide, there may need to be a similar effort to manage the engineered small molecule toxicants that are now ubiquitous in the environment. For example, the 250MM pounds of glyphosate used in the US annually are is found in 86% of  rainwater in the US correlated with negative effects on human and animal health. Furthermore, pesticides are linked to biodiversity collapse of insects (review: Wagner et al 2021, pollinator collapse: DiBartolomeis et al 2019) and amphibians (Brühl et al 2013). There is active debate around whether crops engineered to be resistant to pesticides and herbicides increase or decrease net chemical usage, and at least one study contends that there is a 7% increase of pesticide between 1996 and 2011 (Benbrook 2012). It is essential to develop solutions which simultaneously protect the food harvest while minimizing usage of reactive small molecules that disrupt the environment. Solutions may include plant engineering, symbiote engineering or bioorthogonal chemistry innovation.

Today: ~2 pounds per acre of pesticide for corn, primarily acetochlor and glyphosate (using data from Minnesota in 2015 as an representative example). 

Target: Anything less than today’s level of pesticide usage, with less associated environmental off-target effects that will only be a non-negative impact to the farmer’s yield, could be a significant contribution.

Energy 

Green Hydrogen Cost ($/kg)

Why this matters: Hydrogen is a store of energy that can be created and utilized by biology and machinery alike. While Hydrogen has many challenging aspects including storage, explosion risk and leakages which can increase radiative forcing (Ocko et al 2022), hydrogen is still an important frontier with significant potential in a clean energy future. Even if hydrogen is not an ideal energy source, it could at least be an important energy carrier (MacKay 2008, page 129). The Biden Administration has utilized the Defense Production Act to utilize federal resources in order to rapidly increase clean hydrogen capacity in the US. For a review of the present and future of Green Hydrogen’s potential, we direct the reader to the 2021 Report from Columbia’s Center on Global Energy Policy.  The NREL Hydrogen Production and Delivery website is a good high-level introduction to the technical challenges such as maximizing algal hydrogen production and decreasing hydrogenase oxygen sensitivity (eg, Swanson et al 2015). It’s possible that ambitious new approaches of de novo protein engineering, algal production or metagenomic searches might yield new paths to progress. 

Today: $5/kg (source: KPMG)

Target: $1/kg, set as an “Earthshot” by the US Department of Energy.

Sustainable Aviation Fuel Cost ($/gallon) 

Why this matters: Planes need to fly to maintain the global economy, and aviation is 2.4% of the global CO2 emissions. Sustainable aviation fuels (SAF), which could be an 80% reduction in lifecycle GHG emissions , are one path to decarbonize the aviation industry: IATA thinks there could be 2% market penetration within 5 years (IATA.org factsheet). According to the Department of Energy Biotechnology Office, the US produces one billion tons of dry biomass that can theoretically be collected to produce 50-60 billion gallons of biofuels (Rogers et al, 2016), and SAF could be one possible output. The Biden Administration has formalized the opportunity into the Sustainable Aviation Fuels Grand Challenge (fact sheet presentation). 

Today: As of May 2022, a gallon of SAF costs $8.67. Regular jet fuel is about $4.15 per gallon. (Flying Magazine, May 2022)

Target: SAF should be cheaper for the consumer than traditional jet fuel.

Work ahead

It was outside the scope of this first draft to specifically identify and describe the critical subproblems inside each benchmark. One reason subproblems were not carefully explored in this draft is that choice of subproblems is likely based on assumptions (eg, on deployment embodiments, organism choice, priority) that need to be carefully workshopped with stakeholders. 

Within each of these benchmarks is decades of work that each have reached subtle but important bottlenecks that need to be overcome, and an intention of this first draft is to collectively work toward the identification of the subproblems which will be the nucleation point of important innovations.

For now, in the spirit of bottom-up innovation, the discovery of subproblems is left as an exercise to the reader. If you want to start building technologies that work towards breakthroughs in these benchmarks, reach out to me on Twitter or please comment here in this document.

Acknowledgements: Many amazing people have made contributions to this document. As this is a dynamic effort, with surely some errors in its current form, I am intentionally not sharing the list of acknowledgements today. Those who have contributed so far, thank you so much! Any errors or misjudgments published here are solely my responsibility.

Last week I wrote about early stage funding in science (and thank you to the community for all the input). Yet proper funding is only one part of a productive community, what are the other parts? This week I want to explore illustrative successes of funding, shared resources and shared challenges.

Since graduating college in 2008, I've observed first-hand the growth of three hyperproductive communities: computer science, biotech and brazilian jiu jitsu (BJJ). Now, I know BJJ is a weird one and I’m not here to bro out:  BJJ’s growth and technical advancement is exceptional (more so than other trendy sports such as climbing and surfing), and I contend it instructively fits the model. 

Computer Science needs the least introduction of how transformative it has been to our world. We all carry our smartphones and interact with apps that have been optimized for over a decade of shared ideas and best practices: that world has been built by hustling entrepreneurs that extend ideas from academia and industry in pursuit of the fabled riches. We can search for loosely related phrases to get us the result we want (NLP), enjoy selfie filters with cat ears that seamlessly track our heads (computer vision), interact with a new internet of money (cryptography) and even rely on progressively more reliable self-driving cars (reinforcement learning).  But, from the practitioners’ perspective, what are the characteristics that made it so hyper productive?

+ Common infrastructure for experimentation: Anybody with a minimally powered laptop could start programming, and the dominant languages (eg, python) created package managers to rapidly load in the work of other people.  

+ Common infrastructure for scaling: The iPhone, Facebook and AWS all were game-changers because they were paths to infinite scale. If you made something that people liked, you could go to a million users in days (eg, the Shotgun App in 2009) or could scale your computation beyond any physical limits of computer hardware (eg, MapReduce in 2004). Since 2009, we take for granted that a software idea that works for 10 people can readily scale to 10 Million: this certainty of the potential to scale was essential for the intense ambition that fueled the rise of software eating the world.

+ Common goals: Computer science has been rich with shared public challenges (eg, the $1MM Netflix Challenge in 2006) and dataset-driven competitive benchmarks (eg ImageNet). These public goals act as lenses to focus effort and surface talent. Furthermore, when transformative ideas are invented, the community can immediately appreciate the value.

+ Enforceable honesty: The community can immediately test the validity of your claims and ideas. Because of the common infrastructure and community practice of open source, when big ideas come along they can be replicated and validated. AlexNet took the computer vision community by storm in 2012 with their use of deep neural networks on the ImageNet challenge: the importance of the work was quickly appreciated across all of computer science because anybody could download the code and replicate it themselves.

+ Playbooks for outsized success: Young, ambitious people have clear targets to aim for as they prioritize their time. By 2010, Y Combinator had codified the software startup playbook, created the archetypal software founder and illustrated the incredible riches of those who play the game well. I met a woman at a San Francisco boat party in 2012, in her words: “I’m working at a Series C company backed by a top-tier VC after finishing my undergrad in computer science where I interned at Google and Facebook. In my spare time I’m doing customer discovery in two potential markets [A] and [B] and networking actively for potential co-founders. In two years I’ll apply to YC.” I was super impressed, both by her ambition and the complete clarity of purpose to her fulfilling the archetype of the software entrepreneur of the era.

+ Funding to try things: The amazing thing about computer science is that the marginal cost of an experiment is virtually zero. Beyond student salaries (or the heralded Instant Ramen diet of founders), resources are not really part of the conversation at the early stages of innovation in computer science. All the cloud services are quick to offer free credits and there has been a constant increase of computer science fellowships. 

I chose Professor Fei Fei Li to be in the header image because of her exemplary contributions to computer science. It was her idea to make ImageNet into *the* data benchmark for the field of Computer Vision. But more than a benchmark, it was a training dataset good enough to enable the first practical use of multi-layer neural networks (AlexNet). Similarly, beyond academic contributions of her own, she has been behind startups and AI strategy of the biggest companies in the world. 

I was in the Stanford Computer Science buildings in 2010 (in Fei Fei’s group actually), and I remember the energy in the hallway conversations. Us graduate students would brainstorm how AWS, Facebook minifeed and the iPhone (common infrastructure) could be the platforms to address market opportunities (common goals) via solving tech problems (enforceable honesty). These conversations would have a certain formula of combining a technology challenge and a market opportunity. Eg, 'Ooh we could fuse iphone sensors via ML to determine precise indoor location and share it to twitter!' or 'we could use iphone cameras and computer vision to overlay ratings to store fronts when you walk through palo alto!'. Regardless if those conversations actually converted to startups of any quality, I’d argue that just those brainstorms alone helped computer scientists understand their personal value.

Before computer science matured into the five elements bulleted above, it was the era (early 2000s) in which the stereotypical startup was founded by a fast-talking MBA who then went hunting for programmers as his first hire. But individuals like Zuckerberg, Brin and Page showed us that the best founders are practitioners. And, by the entrepreneurial brainstorming and pitch sparring toward a Y Combinator (or similar) application, the technical graduate students also taught themselves the language of business and techniques for self-advocacy. 

In many ways, biotech has become similar to computer science. The exact same entrepreneurial discussions happen in the conversations of bioengineering hallways today: “Ooh we could use CRISPR to disable this rare, inheritable genetic pathology!” or “Oh we could computationally redesign this protein to cross the blood brain barrier with a small molecule payload!” or “wait, what if we could modulate this obscure metabolic pathway to enable novel combinations of already approved drugs?” These conversations are fun, and I’ve seen first-hand the business sophistication of bioengineering graduate students increase significantly over the past five years.

Biotech has been on its own exponential growth trajectory and only continues to accelerate. Over the past decade it has matured its own creative community  with noteable technical milestones being the sequencing of the human genome (~2001, depending when you call it “done”), next-generation sequencing (2008) and CRISPR (2012, and timeline of discoveries here). I put Professor Jennifer Doudna in the header because the discovery of CRISPR (for which she earned a 2020 Nobel Prize shared with Emmanuelle Charpentier) was an inflection point in biotech’s growth. Suddenly the genome became immediately editable (credit to the work of Professor Feng Zhang and others), and the tools for any scientist could be delivered to their lab via AddGene or IDT overnight. In academia, it means that an animal phenotype could be developed in days rather than years. In industry, it was a foundational change in how the startup ecosystem viewed human health. Billions of dollars of value have since been created, and a new generation of scientists have been developed in a matured entrepreneurial community around human health.

+ Common infrastructure to prototype: Every biolab has E Coli and HEK cells and access to standard data acquisition tools (sequencing, imaging, etc.). Engineered DNA can be delivered within 24 hours for use in experiments. Ideas can be tried within days.

+ Common infrastructure to scale: In 2022, this is much less linear in biology than computer science. While there are cloud labs and CROs, going from milligrams to kilograms of biologics is still a heterogenous process. However, the path to impact in biotech is typically focused around therapeutic milestones in animals then clinical trials, and this is becoming more and more codified. 

+ Common Goals: The big problems of biotech have been mostly implicit to those who know the space: gene editing (eg broadening targetable regions, less toxicity per edit), sequencing (cheaper, faster, longer reads; spatial information preserved, protein sequencing), protein structure (e.g, AlphaFold), delivery (eg, crossing blood brain barrier, targeted cell types). The best, most specific challenges have been at the interface of computer science and biology, eg CASP and CAPRI. While there have been some facilitated competitions of wetlab techniques (eg, CZI’s spatial transcriptomics project), shared explicit goals and challenges have not yet played the same role in bioscience as computer science.

+ Enforceable honesty: The biggest ideas in biotech are the ones that every lab can use. Optogenetics, GCaMP, and CRISPR took the field by storm because they are rapidly distributed, validated and utilized by other labs. It’s important to point out that while peer reviewed science has been a dominant model (imperfect but good), the most “honest” approach is one that lets peers easily try new ideas themselves.

+ Playbook for outsized success: In what I call the “Kendall Square Biotech Model”, there is now a very established path for a precision medicine company: what is your rare monogenic disease (preferably an orphan disease), what is your animal model and what is the next few indications by which you could partner with the existing giants? This relatively homogenous path to impact has been a lens for talent: young ambitious creatives can structure their efforts toward work that could fit this model. Those that succeed are handsomely rewarded.

+ Funding: The marginal cost per experiment in biotech is considerable. As a ballpark number, consider that really trying an idea might cost $5-$50k in reagents, equipment and animals. Consider further that the cost of the journey to a human therapy is in the hundreds of millions. However, to meet this challenge and chase the opportunities exemplified by the highly successful startups (Editas, Beam, Moderna, etc.) there has been rapid growth in many new types of investment vehicles, PhD fellowships, grants and sponsored research agreements. 

Personally, I’m extremely excited by bioengineering and would like to see this approach flourish in domains outside the Kendall Square Biotech model. 

So, why Jiu Jitsu?

The past 15 years have seen an explosion of new fitness outlets: crossfit, climbing, MMA, boxing, etc.. I’ve done all of these and I love them, and one could make a good argument that climbing has seen incredible amount of industrial growth. But I consider BJJ to be special because of both the industrial growth (thousands of people are now able to make careers from it) and the technical growth of the field. Simply put, the modern winning techniques are almost unrecognizable from just a decade ago, and this is because of a combination of the nature of the sport and the incentives in the community.  

Consider one illustrative example: In 2019 a relatively unknown fighter from Australia named Lachlan Giles stepped onto the biggest stage of the world (ADCC) and earned a bronze medal, winning his matches in shockingly fast time (about 5 minutes out of the 30 minute possible combined match time). The reason this turned so many heads is that Giles only weighs 160lbs and he was fighting in the open weight division, meaning the three people he beat were all heavyweight champions who weighed ~220lbs each. These unexpected victories happened because Giles is a very intelligent fighter who devised a set of techniques (variations on the 50/50 position to heel hook) few were prepared for, and he won his matches with his invention. 

“It’s a sign of how amateur our sport is. Imagine a 160lb olympic wrestler competing against a 260lb olympic wrestler. …It’s just insane” Said competitor Gordon Ryan, famously snarky, as he watched Giles’ matches unfold, and then went on to beat Giles and for the gold medal minutes later. [note that Ryan is coached by possibly the greatest single mind in martial arts, John Danaher].  

Just visually speaking, the technique of the sport has been turned on its head. How Giles won his matches looks nothing like how people won matches a decade ago: one looks like wrestling and the other has no real comparison. Through knowledge exchange and the community testing different approaches in competition, the sport shifted from a focus on upper body techniques (chokes, arm locks) to lower body techniques (leg locks, foot locks). When interviewed after winning his match for a bronze medal, Giles was asked, “you’re now a superstar, what do you think this will do for your life?” “Well, I hope it’ll sell more instructionals!”

Instructionals are the heart of why I think BJJ has something to show us in science+technology, and why I chose to put a photo of Professor Bernardo Faria as the header image. As a former champion himself, he converted into a content creator for BJJ techniques, then created BJJFanatics, the leading platform for other champions to create+distribute paid instructionals. The upside for the inventors of the best techniques now get paid millions of dollar per year as customers pay up to $100 for a few hours of recorded online lecture. Or, in the case of one notable athlete who had lived in astute poverty to pursue his dreams in the sport, Andrew Wiltse was able to move out of his homemade shed stapled to the gym.

So how does this map to the dimensions from CS and biotech? 

+ Common infrastructure to prototype: Every hobbyist has mats or a grass field to try things on their friends.

+ Common infrastructure to scale: Competitions are everywhere now and YouTube is surfacing new talents every day. Top competitors will do seminar circuits which are limited, but now with BJJFanatics, the best ideas+teachers can infinitely scale their reach across the market.

+ Common Goals: While there are the biggest stages at ADCC, IBJJF, there are many rapidly rising shows which showcase such as EBI, WNO and fun themed events like High Rollerz series.

+ Enforceable honesty: The martial art of jiu jitsu was designed to be practiced at full strength and full speed, which is essential for the rate of innovation for ideas that work in competition. Even other spar-centric arts like boxing or kickboxing couldn’t test out new head kicks on their friends, the brain damage would be too great from full strength prototyping. People can then watch the new techniques on youtube or BJJFanatics, take them to their local gym, and validate the results for themselves.  

+ Playbook for outsized success: There are now fairly clear milestones to make a successful career in this sport with everybody following the success of most exemplary winners (eg, Gordon Ryan, Craig Jones). Winning large tournaments with novel approaches is a sure way to be invited to make a BJJFanatics instructional. Additionally, there is innovation around sharing a gym’s philosophy and approaches, which can build highly engaged audiences (eg, the Daisy Fresh gym in rural Illinois hav).

+ Funding: The marginal cost to run experiments is zero, and the amount of sponsorships is growing rapidly. This is creating crops of new talents that have have been training the sport since elementary school (eg, Grace Gundrum, the Ruotolo Brothers, Cole Abate) that bring new audiences and investment into the community.

So what does this mean for climate-positive science and technology? It means we have a lot of work ahead of us, and drilling into these aspects of the creative community will be topics for future writings. I’ll leave it this week with some open questions for these different dimensions.

+ Common infrastructure to prototype: Do we have any? Do energy projects have a common toolkit to build from (ie, is there a common “green hydrogen” reactor)?

+ Common infrastructure to scale: For specific verticals inside “climate positive” efforts, which ones have the the most linear paths to scale?

+ Common Goals: Everybody talks about “Net Zero,” but I contend that’s too broad to be actionable inside the community (Aside: I personally don’t find “Net Zero” very exciting, how do we brand it as a more optimistic, forward looking framing as technologies we would want anyway). For example, in the carbon capture community, gigaton carbon capture at less than $100/ton is an unofficial goal. What other ones are there?

+ Enforceable honesty: How do we know a good idea claim from a bad idea claim? How do we quickly spread, test and refine ideas?

+ Playbook for outsized success: What are the biggest wins so far? Are there common pathways we can learn from?

+ Funding: Assume that experiments cost at least as much as biotech experiments, say $5-$50k marginal cost to really try a new idea. Early stage funding was explored last week: lots of startup money, lots of scale money, but not a lot of rapid, early-stage science support.

Thoughts/ideas/corrections? Please let me know in the comments or on Twitter

Update 02/01/2022: Thank you everybody who contributed to this twitter discussion. I have updated the collection of funding agencies for early-stage science funding, and should also clarify the goal of this post is to focus on pre-IP lab-stage efforts, which is out of scope for many private investors and leans more toward philanthropy or government. For an overview on the climate capital stack, ClimateTechVC has one of my favorite write-ups.

Doing what we already know how to do takes the world from 1 to n, adding more of something familiar. But every time we create something new, we go from 0 to 1. The act of creation is singular, as is the moment of creation, and the result is something fresh and strange.
Unless they invest in the difficult task of creating new things, American companies [and Science] will fail in the future no matter how big their profits remain today. What happens when we’ve gained everything to be had from fine-tuning the old lines of business that we’ve inherited? Unlikely as it sounds, the answer threatens to be far worse than the crisis of 2008. Today’s “best practices” lead to dead ends; the best paths are new and untried.
-Peter Thiel and Blake Masters, “Zero to One”

There is LOTS of amazing activity around new models for funding and doing science. For the fastest overview, I suggest you bookmark Sam Arbesman’s Overedge Catalog. The newest organizations are mostly focused on translational computer science and biotech, especially for those funding vehicles for the earliest stage of research. This creates focus and community around topics like AI and Longevity (both great topics), and new models for producing biomedical startups. But where is the funding for early stage science that produce positive environmental impact?

From my experience, it seems that while speculative/ambitious science is being supported by new capital sources in many areas, early stage “climate-positive” efforts still seem to be stuck in older, slower models of funding. I’m writing this post to ask the community for input, and can share my own journey to explain my perspective. If I’m missing anything please comment here or ping me on Twitter

It was the fourth year of my PhD in bio/neurotech when I first gave even the slightest consideration to working on climate tech. Bluntly, I’d learned to ignore the hyperbolic rhetoric and photos of dead polar bears: it was all positioned as hopeless or purely political, plus I had my own ambitions and career to think about. But being exposed to the domain of carbon capture technology in 2018 by my friend Sarah Sclarsic was a gateway drug of sorts: it re-framed “climate” as a set of technological challenges, and tech challenges is what I do. So where do I start?

It’s both a cliche and truism in science that asking the right question is the most important thing you will do. Personally, I’ve observed the route to the right question is itself a series of trying things, just like an testing any other hypothesis. So under the context of a class project, I did a deep dive on how I, as a former software entrepreneur, would professionally evaluate the space of carbon capture. That 2019 report is here, and I was searching for order-of-magnitude areas for improvement that might make carbon capture either sufficiently profitable to grow as a startup or radically scalable to immediately matter on the global scale. (Aside: if you want to go deep on carbon capture, the CDRPrimer is an excellent starting point). As I saw the space of carbon capture startups proliferate rapidly (Airminers has a great running list of carbon capture companies), I decided to focus toward the other extreme: searching for the largest capacity carbon capture approaches. 

The two biggest global levers we have to get carbon dioxide out of the atmosphere is through the ocean (eg, alkalinity enhancement) or rock weathering. If you focus purely on the drawdown potential and disregard a need to produce a marketable good, the theoretical annual carbon capture capacity from the ocean and the rocks is many gigatons per year, which is the scale that humanity will need by 2030. At this point, we now had developed a nucleus of bioengineers working together, and collectively we chose to focus on rock weathering because it felt the most addressable with our skill sets. We know microbes can eat rocks, how might we harness nature’s methods toward a 1000x speed up the natural rock weathering process? 

To really approach the challenge of microbially-enhanced rock weathering, we knew as a team we needed both the idea and the funding to let us pursue that idea. Paul Reginato and I were the two grad students working most on this exploration. We were lucky to be mentored by Professors Ed Boyden and George Church who had built high-profile labs capable of allowing their graduate students to freely pursue such ambitious yet orthogonal goals to the lab’s prior body of work. And as we were still searching the surface of approaches, the MIT Climate Grand Challenge (CGC) was announced. 

In my opinion, the MIT CGC was the most under-appreciated funding innovation of the past two years. As has been well documented by several other sources, there has been a LOT of funding innovations for science recently, and this is all very good. For example, Fast Grants rightfully received a lot of attention for their simple model for speed and success at encouraging ambitious science projects. I think the CGC got less attention because it was MIT-specific, and admittedly, it’s still too early to tell if the model was a success. The CGC design was $100k for 2-page letters of intent to fund the production of a 25-page white paper. In principle, this white paper would then be used by the MIT Development Office to fundraise for the university and to then support the execution of the ideas in that white paper. But from our experience as recipients of this support, it was much more: it was a catalyst to meet many highly relevant researchers.

To mechanistically study the conversion of silicates to carbonates (rock weathering),  we needed to collaborate with chemists and geologists. One of the good design aspects of the CGC was a series of events in which accepted LoIs were presented in short video lectures. I went to the majority of these lecture series, and reached out to all the professors that were doing adjacent projects. This turned into friendships and collaborations with Admir Masic (Civil Engineering), Yogesh Surendranath (Chemistry), Matej Pec (Geology) and Loren Looger (UCSD/HHMI Bioengineering) and many others.  I am confident to say that I don’t think we all would have connected if it were not for the forcing function of the white paper: it was both financial freedom to experiment and an expectation of a final deliverable. Furthermore, the money and the deadline of a deliverable was the nudge for me make space in my schedule of other PhD commitments and start getting some intuition of what it means to convert silicates to carbonates.

My experiences at the end of graduate school have made me deeply curious about the onramp for people like me to translate their abilities into climate-positive efforts. From my perspective, the ideal catalyst would have the following features:

• Financial security: Let my creativity be the limiting factor to the experiments, and for my salary to be covered to give me the time to think clearly. As a semi-arbitrary number, I’d posit that $100k is the minimal amount to fund 1-3 people and their experiments.

• Encourage ambitious, early science: Fund me at the idea stage, before I have a publishable result or an established record in the field. Talk to any scientist about applying for NIH/NSF funding and you’ll hear a snarky quip to the tune of “they’ll fund me for what I’ve already done.”

• Team building and community membership: Break me out of my own head, let me build and fund my own team for the short period it takes to test the earliest stages of an idea. Fellowships are a leading model for early stage science for good reason, but one limiting factor is that they only support one person. Ideally, the other people supported by the catalyst would be great candidate teammates or at least help me grow my collaborator network.

• Bridge to the next step or freedom to move on: If my first efforts succeed, how can I be confident that there is the next stage of support for that project? And, if my effort fails for the right reason (eg: my work uncovers an unexpected blocker or the idea simply doesn’t seem to well enough to continue), can I still be recognized for it? The lack of a place to productively share good efforts that reach a noble dead end in science has been long lamented.

• Soft guidance toward the most direct path to impact: Help me confirm that I am working on the right problem from the right angle. There is a productive tension in science, especially climate, between “bottom-up” and “top-down” efforts. Top-down efforts are informed by first principles and roadmapping-based exercises to break a big problem into actionable chunks that can be orchestrated amongst practitioners. In contrast, bottom-up ideas come directly from the practitioners who have their own perspective and informed ideas. And while I think the majority of breakthroughs may well be accomplished from top-down science (note: top-down efforts are not without their own failure modes), it would be hubristic to think that all good ideas could be covered by a roadmap. One could make an argument that the true paradigm changers could not be covered by a top-down effort by definition.  In practice, all efforts will sit somewhere on the spectrum of bottom-up vs top-down: even bottom-up efforts would still require some scaffold of ideas, relevant efforts and prior knowledge. Mentors and community must be there to reverse engineer the maximum possible impact of an idea.

So with these desired features in mind, what are the funding opportunities that exist for early stage ideas by early stage researchers? Ignoring university-specific funds (which unless you’re MIT, Harvard or Stanford are probably well under $100k), the options are listed below. Here is a spreadsheet of the granting agencies that support early stage research to earlier career scientists. Please add in rows if I’m missing anything or incorrect (I will do my best to update this screenshot as input comes in).

The answer that I currently arrive at is that there is not much at the moment, especially for the realm of a bioengineering → climate translation (note that Climate Change AI did grants in 2021 for computer science → climate projects). Contrast this to the Impetus Grants or Altos Labs for Longevity. It feels like there is a gap especially for the exponentially growing reach and impact of biotechnology.

This means the main path for a graduating PhD student to work on early stage science efforts is to moonlight during their PhD, apply for a prestigious postdoctoral fellowship, find a creative+established PI who supports novel ideas or work at a well-capitalized company that is doing R&D in a topic that excites you (eg, Ginkgo spun out Allonia to address pollution). Is this optimal? Are there domains that have done this better that we can learn from?

I wonder if there is more we can do.

See you guys next week or on Twitter.