The Black Box of Big Models, the White Box of the Lab

The black box of large language models drives us mad, so blockchain now tries to install a transparent white box inside the laboratory.

---

1943: A Quantum Cat, A Teenage Prodigy, and the Question “What Is Life?”

In 1943, the owner of the quantum cat, Erwin Schrödinger, delivered a high-wire lecture in Dublin. From the perspective of statistical physics he linked atoms, life, and cells. Across the Atlantic, the 15-year-old James Watson was already a freshman at the University of Chicago. After reading the lecture-turned-book What Is Life?, Watson decided genetics would be his lifelong quest.

Ten years later, PhD in hand, Watson unveiled the double-helix structure of DNA. A 25-year-old Nobel Prize was already waiting on the mantelpiece.

---

Grafting, Cloning, Gene Editing

There are two trees in front of my house; one is a jujube, and the other, too, is a jujube.

Anyone who has sat through middle-school biology now knows that a gene is an information snippet on DNA—like a function in code, the minimal unit of execution. DNA itself is an instantiated module; RNA plays the router and messenger, ferrying instructions to their intended targets.

Watson showed us the structure, but not how to use it. Knowing the cat is in a superposition is easy; quantum communication takes decades. Watson, at least, was luckier than Schrödinger. In the summer of 2012, Emmanuelle Charpentier and Jennifer Doudna discovered that CRISPR sequences and Cas proteins could be paired: snip a chosen locus, insert a desired snippet, and let the cell’s own repair machinery finish the graft—quietly, invisibly.

It is exactly like pruning in a garden: you eyeball the branches, splice them, and, without understanding the underlying botany, still manage to make different plants grow together—trial and error suffices.

Cloning follows the same rulebook: nucleus and cytoplasm can be separated and re-paired. After enough experiments, isomers of organisms appear, as What Is Life? predicted.

Gene editing is merely cloning pushed one notch smaller. From the perspective of atoms, life is just heat death in slow motion—time can stretch or compress but never reverse.

We graft fruit trees, we clone sheep—can we edit humans?

In 2018 the “mad scientist” He Jiankui became either Eve or the serpent. He edited the genomes of twin embryos whose parents carried HIV. Pandora’s box opened. Cloned animals can be euthanized; edited humans—are they still human?

---

The Lure of Longevity

Yet gene-level control exerts a fatal attraction: longevity. Find the snippet that governs lifespan, hack it like a North-Korean cyber-soldier, nudge the value from 100 to ∞. Even adding a single zero would be enough.

---

Coinbase Alumni Turn from Coins to Chromosomes

In 2023 Paradigm co-founder Fred Ehrsam left crypto and founded Nudge Biosciences. Fred also happens to be the co-founder of Coinbase; after the 2017 listing he pivoted to venture capital.

Back in 2017, Paul Kohlhaas was head of BD at ConsenSys. A year later he quit: why not use blockchain for something more interesting—say, science?

In 2018 Molecule was born, one of the first attempts to combine blockchain with biological research. Meanwhile DeepMind’s AlphaFold (2016) was already dazzling structural biologists.

By 2020 AlphaFold2 cracked the protein-folding problem. A 4-year-old algorithm was pre-booking half of the 2024 Nobel Prize in Chemistry.

Fred’s 2023 switch was late; in 2020 Coinbase co-founder Brian Armstrong had already launched ResearchHub to dismantle the university–journal–grant trinity. Universities control tenure, publishers skim profits, reviewers work for free, and grant committees hold the sword of Damocles. Scholars pay to submit, publishers charge to read, and only the middleman wins.

AI, science, and tokens converged on life sciences—biology really was the science of the 21st century.

---

DeSci as the Pharma of Immortality

Decentralized science (DeSci) is pharma R&D in a permissionless wrapper.

It is the crypto remix of AI4Science, laser-focused on AI, longevity, and drug discovery.

Remember Paul Kohlhaas’s Molecule? Balaji invested in 2022—nobody can resist the promise of a longer life.

One step further: in 2022 Paul launched Bio Protocol, intent on giving crypto barons extra decades. Sub-DAOs cover everything from male pattern baldness to toe-nail fungus.

In 2024 a reborn CZ and a youthful Vitalik shared a stage in Bangkok. Vitalik recommended VitaDAO’s supplement VD001; CZ’s YZi Ventures wrote the check. BIO tokens listed on Binance. Paul even cloned PumpFun into PumpScience—meme meets molecule.

But after the pump comes the reckoning. A new drug costs a billion dollars and ten years; the secondary market won’t wait five minutes. Raising money and not pumping the chart is the new mortal sin.

---

Enter the Agents

Yet the story does not end; the Agent wave has arrived. AI agents may finally speed up research. In February 2025 ResearchHub raised a $2 million Boost round; DeSci agents now have peer review.

On August 1, 2025, Bio Protocol released V2: a new launchpad, BioXP points, and BioAgents—built on ElizaOS, forever surfing the latest trend. Within seven days more than 100 million BIO were staked (8,000 million on the 7th alone—data integrity may vary). Tokenomics were redesigned to favor small floats and continuous sponsorship over mercenary dumps.

---

Catching Up with AlphaFold

Still, DeSci lags behind AI4Science. AlphaFold open-sourced its database in 2021; 200 million protein structures later, it has mapped most known species.

Bio Protocol begs the FDA to open pharma vaults; V2 will fast-track drugs in the UAE where human-trial rules are looser. Whether the next hero is Watson or He Jiankui remains to be seen.

---

Epilogue: Resurrection, Evolution, Extinction

GPT-5 may disappoint in the general case, but in high-value verticals—medicine, science—the scaling laws may yet bloom. Colossal Biosciences, deep in Silicon Valley, is already CRISPR-engineering a “woolly mouse” (mammoth genes in a mouse) and a ghost-white dire wolf.

One day humans may evolve.

One day humans may vanish.

The Black Box of Big Models, the White Box of the Lab

The black box of large language models drives us mad, so blockchain now tries to install a transparent white box inside the laboratory.

---

1943: A Quantum Cat, A Teenage Prodigy, and the Question “What Is Life?”

In 1943, the owner of the quantum cat, Erwin Schrödinger, delivered a high-wire lecture in Dublin. From the perspective of statistical physics he linked atoms, life, and cells. Across the Atlantic, the 15-year-old James Watson was already a freshman at the University of Chicago. After reading the lecture-turned-book What Is Life?, Watson decided genetics would be his lifelong quest.

Ten years later, PhD in hand, Watson unveiled the double-helix structure of DNA. A 25-year-old Nobel Prize was already waiting on the mantelpiece.

---

Grafting, Cloning, Gene Editing

There are two trees in front of my house; one is a jujube, and the other, too, is a jujube.

Anyone who has sat through middle-school biology now knows that a gene is an information snippet on DNA—like a function in code, the minimal unit of execution. DNA itself is an instantiated module; RNA plays the router and messenger, ferrying instructions to their intended targets.

Watson showed us the structure, but not how to use it. Knowing the cat is in a superposition is easy; quantum communication takes decades. Watson, at least, was luckier than Schrödinger. In the summer of 2012, Emmanuelle Charpentier and Jennifer Doudna discovered that CRISPR sequences and Cas proteins could be paired: snip a chosen locus, insert a desired snippet, and let the cell’s own repair machinery finish the graft—quietly, invisibly.

It is exactly like pruning in a garden: you eyeball the branches, splice them, and, without understanding the underlying botany, still manage to make different plants grow together—trial and error suffices.

Cloning follows the same rulebook: nucleus and cytoplasm can be separated and re-paired. After enough experiments, isomers of organisms appear, as What Is Life? predicted.

Gene editing is merely cloning pushed one notch smaller. From the perspective of atoms, life is just heat death in slow motion—time can stretch or compress but never reverse.

We graft fruit trees, we clone sheep—can we edit humans?

In 2018 the “mad scientist” He Jiankui became either Eve or the serpent. He edited the genomes of twin embryos whose parents carried HIV. Pandora’s box opened. Cloned animals can be euthanized; edited humans—are they still human?

---

The Lure of Longevity

Yet gene-level control exerts a fatal attraction: longevity. Find the snippet that governs lifespan, hack it like a North-Korean cyber-soldier, nudge the value from 100 to ∞. Even adding a single zero would be enough.

---

Coinbase Alumni Turn from Coins to Chromosomes

In 2023 Paradigm co-founder Fred Ehrsam left crypto and founded Nudge Biosciences. Fred also happens to be the co-founder of Coinbase; after the 2017 listing he pivoted to venture capital.

Back in 2017, Paul Kohlhaas was head of BD at ConsenSys. A year later he quit: why not use blockchain for something more interesting—say, science?

In 2018 Molecule was born, one of the first attempts to combine blockchain with biological research. Meanwhile DeepMind’s AlphaFold (2016) was already dazzling structural biologists.

By 2020 AlphaFold2 cracked the protein-folding problem. A 4-year-old algorithm was pre-booking half of the 2024 Nobel Prize in Chemistry.

Fred’s 2023 switch was late; in 2020 Coinbase co-founder Brian Armstrong had already launched ResearchHub to dismantle the university–journal–grant trinity. Universities control tenure, publishers skim profits, reviewers work for free, and grant committees hold the sword of Damocles. Scholars pay to submit, publishers charge to read, and only the middleman wins.

AI, science, and tokens converged on life sciences—biology really was the science of the 21st century.

---

DeSci as the Pharma of Immortality

Decentralized science (DeSci) is pharma R&D in a permissionless wrapper.

It is the crypto remix of AI4Science, laser-focused on AI, longevity, and drug discovery.

Remember Paul Kohlhaas’s Molecule? Balaji invested in 2022—nobody can resist the promise of a longer life.

One step further: in 2022 Paul launched Bio Protocol, intent on giving crypto barons extra decades. Sub-DAOs cover everything from male pattern baldness to toe-nail fungus.

In 2024 a reborn CZ and a youthful Vitalik shared a stage in Bangkok. Vitalik recommended VitaDAO’s supplement VD001; CZ’s YZi Ventures wrote the check. BIO tokens listed on Binance. Paul even cloned PumpFun into PumpScience—meme meets molecule.

But after the pump comes the reckoning. A new drug costs a billion dollars and ten years; the secondary market won’t wait five minutes. Raising money and not pumping the chart is the new mortal sin.

---

Enter the Agents

Yet the story does not end; the Agent wave has arrived. AI agents may finally speed up research. In February 2025 ResearchHub raised a $2 million Boost round; DeSci agents now have peer review.

On August 1, 2025, Bio Protocol released V2: a new launchpad, BioXP points, and BioAgents—built on ElizaOS, forever surfing the latest trend. Within seven days more than 100 million BIO were staked (8,000 million on the 7th alone—data integrity may vary). Tokenomics were redesigned to favor small floats and continuous sponsorship over mercenary dumps.

---

Catching Up with AlphaFold

Still, DeSci lags behind AI4Science. AlphaFold open-sourced its database in 2021; 200 million protein structures later, it has mapped most known species.

Bio Protocol begs the FDA to open pharma vaults; V2 will fast-track drugs in the UAE where human-trial rules are looser. Whether the next hero is Watson or He Jiankui remains to be seen.

---

Epilogue: Resurrection, Evolution, Extinction

GPT-5 may disappoint in the general case, but in high-value verticals—medicine, science—the scaling laws may yet bloom. Colossal Biosciences, deep in Silicon Valley, is already CRISPR-engineering a “woolly mouse” (mammoth genes in a mouse) and a ghost-white dire wolf.

One day humans may evolve.

One day humans may vanish.