AI Got Good at Language. Now It’s Learning the Language of Life. (Eric Nguyen, Co-Founder and CEO of Radical Numerics)
The GeneralistShow Notes
Eric Nguyen is the co-founder and CEO of Radical Numerics, an AI research lab that has raised $50 million to train models directly on biological data. Before starting the company, Eric helped develop Evo and Evo 2, large-scale genome language models trained on unlabeled DNA sequences. Radical Numerics is now building models that can connect information across DNA, RNA, proteins, epigenetics, and other parts of biology, rather than treating each as a separate problem. Researchers have already used Evo to generate viable bacteriophage genomes, and Eric says Radical Numerics’ newer model, Omnii, matched key findings from two years of Alzheimer’s wet-lab research in a matter of days. He also believes these tools could make it easier to create dangerous pathogens, which is why the company is working on both biological design and biodefense.
In our conversation, we explore: What AI models can learn by treating DNA as a language Why reading scientific papers is not the same as learning directly from biological data How Eric’s unusually free-range childhood shaped the way he follows his curiosity Why biology may have more useful data than researchers know how to use How Radical Numerics plans to connect information across DNA, RNA, proteins, and other biological systems Where the company sees early opportunities in drug discovery, diagnostics, synthetic biology, and biodefense Why testing AI-generated biology in the lab is still slow and difficult How models that design biological systems could also help detect dangerous or manipulated pathogens How to make powerful biology models safer without eliminating the capabilities that make them valuable — Thank you to the partners who make this possible Ahrefs Brand Radar: Find your brand in AI results. Brex: The intelligent finance platform. Guru: The AI source of truth for work. — Timestamps (00:00) Intro (03:35) An overview of Radical Numerics (06:35) From protein models to modeling all of biology (11:08) Why they started with DNA (15:04) The process of mapping DNA as a language (19:47) What’s unknown, and how we learn from novelty (26:24) The limits of language models in biology (31:15) Eric’s free-range upbringing and path to his PhD program (41:20) Applying long-context models to DNA and meeting his co-founders (46:36) Biology’s untapped data opportunity (49:02) Why biology needs multimodal AI (55:30) How better general LLMs benefit Radical Numerics (57:19) The challenges of biological verification (1:02:05) Making biology more concrete (1:04:51) Radical Numerics’ strategy and early use cases (1:07:26) Balancing safety with capable AI models (1:15:47) What success in biodefense looks like (1:18:09) Final meditations — Follow Eric Nguyen LinkedIn: https://www.linkedin.com/in/nguyenstanford X: https://x.com/exnx Website: https://erictnguyen.com — Resources and episode mentions: https://www.generalist.com/p/ai-got-good-at-language-now-its-learning — Production and marketing by penname.co. For inquiries about sponsoring the podcast, email [email protected].
More from The Generalist
Pete Florence and I began by agreeing that his company has an excellent name. Generalist builds AI for robots, including models that can begin learning a new task from a single demonstration. We talked about dexterity, the strange business of gathering physical experience for a machine, and what we might build if robots could do much more of the work. — Every timestamp 00:00 Intro 02:24 GEN-1.5 and one-shot learning 08:55 What makes a task difficult 10:53 Dexterity in robotics 14:53 Matt Mason’s definition of a hand 17:34 Giving robots a sense of touch 20:30 Growing up in a family of technologists and scientists 22:46 What drew Pete into robotics 27:38 Generalist’s founding thesis and what sets it apart 32:17 Protecting simplicity 35:01 How Generalist’s data collection has evolved 35:30 Generalist’s scaling philosophy 39:53 Generalist’s build-versus-buy philosophy 42:15 Balancing general intelligence with commercial deployment 44:24 Why Pete is betting on dexterity 47:33 The durability of the robotics application layer 49:09 Open-weight robotics models and data flywheels 50:52 Geopolitics, safety, and alignment 52:42 The parallels between teaching humans and robots 1:00:05 The timeline for robotics progress and adoption 1:03:37 Book recommendations — Transcript: https://www.generalist.com/i/220756708/transcript — Pete’s complete reading list The Score Takes Care of Itself: My Philosophy of Leadership: https://www.amazon.com/Score-Takes-Care-Itself-Philosophy/dp/1591843472 Wooden on Leadership: How to Create a Winning Organization: https://www.amazon.com/Wooden-Leadership-Create-Winning-Organization/dp/0071453393 — Other resources Generalist: https://generalistai.com GEN-1.5: https://generalistai.com/blog/gen-1.5 Proprioception: https://en.wikipedia.org/wiki/Proprioception Cutaneous receptor: https://en.wikipedia.org/wiki/Cutaneous_receptor Fukushima nuclear accident: https://en.wikipedia.org/wiki/Fukushima_nuclear_accident The bitter lesson: https://www.cs.utexas.edu/~eunsol/courses/data/bitter_lesson.pdf GEN-1: https://generalistai.com/blog/gen-1 — Where to find Pete Florence LinkedIn: https://www.linkedin.com/in/peteflorence X: https://x.com/peteflorence Website: https://www.peteflorence.com — Production and marketing by penname.co. For inquiries about sponsoring the podcast, email [email protected].
Jaan Tallinn is a founding engineer of Skype and Kazaa, and a co-founder of the Cambridge Centre for the Study of Existential Risk and Future of Life Institute. For more than fifteen years, he has been among the most consistent voices arguing that advanced AI poses an extinction-level risk to humanity.
Liam Fedus and Dogus Cubuk are the co-founders and co-CEOs of Periodic Labs, a startup building “synthesis superintelligence.” In practice, that means an AI system that closes the loop between hypothesis, experiment, and learning.
Scott Morton is the founder and CEO of Revel, a software platform for testing and controlling complex hardware systems. Before founding the company in 2024, Scott spent nearly a decade at SpaceX building control software for Falcon 9 and Starship. That experience helped him recognize a broader problem: while rockets, nuclear systems, supersonic aircraft, and other advanced machines have grown increasingly sophisticated, the software used to test and control them is often fragmented and decades old. Revel was created to close that gap. The company has raised $180 million, was reportedly valued at just over $1 billion, and counts Impulse Space and Radiant Nuclear among its customers. Its platform scales from small benchtop tests to industrial systems with hundreds of thousands of telemetry channels. Revel has also created its own programming language, RevelCode, designed to combine performance and accessibility with the runtime safety required for high-stakes physical systems.
Farooq Malik is the co-founder and CEO of Rain, a financial infrastructure company powered by stablecoins. Growing up in an immigrant family, Farooq saw firsthand the friction involved in moving money across borders. Alongside co-founder Charles Yoo-Naut, Farooq spent years building infrastructure before stablecoins became mainstream, betting that tokenized money would eventually become a foundational layer of the global financial system. Today, Rain powers card issuance, payments, and other financial products built on stablecoin rails, helping companies move money faster and operate across markets.