Charlie Munger on the origins of Chinese-Americans 🇺🇸 🇨🇳
“The Chinese first came in USA trying to build the Sierra, trans-continental railroad in the winter.”
“Our people were dying and it was just impossible, so they brought in 50,000 Chinese coolies, who were in those days practically slaves.”
“They took the coolies in the mountains and said - you build the railroads and they did it! The Americans couldn’t do it by themselves.”
“Fade in fade out 150 years later, due to immigration, these asians have rapidly become Doctors, Lawyers, Businessmen and succeeded mightily.”
“Every instrument that’s hard to play in symphony orchestra, is played by a Chinese face.”
- Charlie Munger. 2019
Thrilled to share our work on novo protein LYTAC(pLYTAC)/EndoTag in @nature! Using de novo protein design, we created endocytosis triggering binding proteins (EndoTags) to degrade extracellular targets and amplify receptor signaling. https://t.co/lN8NdT8904
We've made a breakthrough in self-evolving AI scientists moving from "search" to "principled discovery": Scientific discovery requires that the search space itself changes, and an AI scientist must perceive this shift without intervention. We built an AI that achieves this for the first time with the ability to discover the scientific vocabulary it reasons in. Evidence, tools, artifacts, verifiers, failures & claims become typed provenance. We show three distinct modalities: 1) retrieval, adding known objects; 2) search, exploring a fixed schema; and critically: 3) discovery, a verified regime transition.
We solve the open-endedness evaluation problem by lifting agentic workflows into a typed copresheaf and proving, via a Kan obstruction, that true discovery is not unbounded generation but a verifiable schema expansion: old evidence is transported by Left Kan extension, and genuine novelty is mathematically quantified by the pointwise residual beyond the transported image - separating discovery from mere search and making novelty objective and measurable rather than a subjective judgment or benchmark delta.
Our AI scientist is built in a way that does not pre-conceive the approach it chooses; instead, we endow the system with formal power to adapt, evolve, and reason from first principles. Case studies include:
1⃣Builder/Breaker model that discovers mode-conditioned compliance in proteins;
2⃣CategoryScienceClaw that finds anisotropic fiber-network stiffness rules.
Great work in collaboration with my graduate student @fwang108_@MITdeptofBE
F.Y. Wang & M.J. Buehler, Self-Revising Discovery Systems for Science: A Categorical Framework for Agentic Artificial Intelligence, arXiv:2606.01444, 2026
Two of the most powerful technologies we've ever known—CRISPR human embryo editing and self-improving A.I.—are getting actualized, and carry known and unforeseen risks
https://t.co/yQY4KqCIhy
https://t.co/08ZSvxZ07k @AnthropicAI
Last week, I was invited to present at @AnthropicAI'e Boston Tech Week event.
Since the Claude Code moment at the end of last year, the way we do science at @ManifoldBio made a quantum leap overnight. But really, it just came full circle.
Manifold was founded by “hybrid scientists” and most of our first few hires could and did both code and do wet lab experiments. This was critical to invent and build the foundational molecular barcoding tech that lets us do the million-scale experiments no one else in the world can do, including interrogating drug candidates directly in vivo.
Soon, the form of most of the data generated at Manifold became next-gen DNA sequencing data. (At the Anthropic forum, I live vibe-analyzed a summary that showed we’ve now done 525 NGS runs generating over 70 tera-bases — 10,000 human genomes worth).
Engineering biology through this NGS lens is both a feature and a challenge. A feature, because this is how Manifold unlocks massive parallelism (a.k.a. GPU-ification of biology). A challenge, because for many scientists this was the first time they couldn’t easily analyze their own data, and became bottlenecked by a dedicated computational counterpart to help turn around insights, which still took weeks.
Over night, that bottleneck has evaporated, aided by a strong data integration layer and an agentic interface we’ve built on top.
Once again, everyone at Manifold Bio has become a hybrid scientist, just in time as the in vivo engine has achieved both scale and richness that have not been possible before.
Even before this, we had already started making immensely valuable discoveries, including shuttles exploiting novel portals to deliver medicines to the brain – a high value problem that led to our first landmark deal with Roche last fall.
Now that the loop is closed for scientists (and agents) at Manifold Bio, the pace of these discoveries is accelerating and we’re about to sweep through many more of the grand challenges in medicine.
Thanks again to the Anthropic team for the opportunity.
Last week, I was invited to present at @AnthropicAI'e Boston Tech Week event.
Since the Claude Code moment at the end of last year, the way we do science at @ManifoldBio made a quantum leap overnight. But really, it just came full circle.
Manifold was founded by “hybrid scientists” and most of our first few hires could and did both code and do wet lab experiments. This was critical to invent and build the foundational molecular barcoding tech that lets us do the million-scale experiments no one else in the world can do, including interrogating drug candidates directly in vivo.
Soon, the form of most of the data generated at Manifold became next-gen DNA sequencing data. (At the Anthropic forum, I live vibe-analyzed a summary that showed we’ve now done 525 NGS runs generating over 70 tera-bases — 10,000 human genomes worth).
Engineering biology through this NGS lens is both a feature and a challenge. A feature, because this is how Manifold unlocks massive parallelism (a.k.a. GPU-ification of biology). A challenge, because for many scientists this was the first time they couldn’t easily analyze their own data, and became bottlenecked by a dedicated computational counterpart to help turn around insights, which still took weeks.
Over night, that bottleneck has evaporated, aided by a strong data integration layer and an agentic interface we’ve built on top.
Once again, everyone at Manifold Bio has become a hybrid scientist, just in time as the in vivo engine has achieved both scale and richness that have not been possible before.
Even before this, we had already started making immensely valuable discoveries, including shuttles exploiting novel portals to deliver medicines to the brain – a high value problem that led to our first landmark deal with Roche last fall.
Now that the loop is closed for scientists (and agents) at Manifold Bio, the pace of these discoveries is accelerating and we’re about to sweep through many more of the grand challenges in medicine.
Thanks again to the Anthropic team for the opportunity.
French founders went from 3.5% of my Y Combinator batch to 20% of the latest one. In 2 years.
Same country. 70 million people. A 6x jump in the French share in two years.
This isn't luck.
The French education system produces generalists with brutally strong hard science foundations (maths, physics, engineering) who can drop into any new field and rebuild it from first principles.
A growing number of them have understood one thing: if you want to build something big, you come to the US and you build it here.
Two things I'd tell anyone French and talented reading this:
Apply to YC. Seriously. There is no downside. Worst case you lose an afternoon on the application. Best case it changes your life. And you can always say no.
It has never been easier to turn an idea into a real company. That's exactly what I'm building with @NanoCorpHQ : you describe the company you want, and you launch it from a single prompt.
The barrier to building used to be technical. Now it's just agency.
Most AI bio companies build their own drugs. We made the opposite bet: build the design engine and sell access, so the best pharma can build better medicines.
We're building the Cadence Design Systems for biology, and hiring on product + platform eng. DM me!
ESMC didn't learn protein biology from a textbook. It learned from 2.8 billion sequences—the full evolutionary record of what works in nature. That's what a world model of protein biology looks like.
Download the model and start building: https://t.co/FQ9JObZv6F
https://t.co/3BWIq8mv2F
A Framework for Autonomous AI-Driven Drug Discovery
Douglas W. Selinger, Timothy R. Wall, Eleni Stylianou, Ehab M. Khalil, Jedidiah Gaetz, Oren Levy
@plexresearch
just look at the timeline chad
- demis hassabis, deepmind → isomorphic labs, $2.7B
- brian, coinbase → newlimit, $3.1B
- sama, openai → retro biosciences, $1.2B
- jeff the chad from amazon → altos labs, $3B
- larry, oracle → $430M into aging research
- jensen from nvidia → backing programmable biology
- dario, anthropic → acquired coefficient bio, $400M
the most successful builders of the digital age are all being pulled toward the next biggest frontier, biology
aging is basically an engineering challenge
bio/acc.
1/n What fraction of the human genome is essential for cellular viability?
Excited to share our preprint that explores this question by combining an unusual CRISPR system, phage promoters, and thousands of deletion launchpads.
@sudpinglay@JShendure
https://t.co/rTO1MyFv3q
A very impressive study for how we could prevent lung cancer more than 5 years before it is diagnosed. Using machine learning, discovery of a 14-plasma protein signature of risk that predicts responsiveness to an antibody therapy to interleukin, IL-1β
Validated across 8 cohorts
@CellCellPress@CharlesSwanton
https://t.co/qpPtgs1dH0