The Generalist

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 jordan@penname.co.

Matan Grinberg is the co-founder and CEO of Factory, an AI company valued at $1.5 billion that helps enterprises like Nvidia, Morgan Stanley, and Adobe automate software development through “Droids,” intelligent agents designed to streamline software engineering. Before Factory, Matan spent more than a decade in theoretical physics, studying string theory at Princeton and UC Berkeley. His work now centers on a different kind of complex system: how software gets built in an era of increasingly capable AI agents, open models, and shifting compute economics.

In our conversation, we explore:
How Emmy Noether’s theorem continues to shape Matan’s approach to technology, business, and AI
Why Matan believes there will always be more problems to solve, even as AI becomes more capable
The resource allocation problem facing CEOs as they balance headcount, compute, and token budgets
Why Factory is betting on model independence and Matan’s take on the SpaceX-Cursor deal
Why Matan pushes back on conflating open models with “Chinese models” and wants a stronger open-model ecosystem
The identity crisis that followed Matan’s decision to leave physics
Lessons from Factory’s first few years, including learning to push back and identify gaps in his own knowledge
Factory’s culture, values, and Matan’s partnership with co-founder Eno Reyes
—
Thank you to the partners who make this possible
.tech domains: An identity for builders at their core.
Brex: The intelligent finance platform.
Persona: Trusted identity verification for any use case.
—
Transcript: https://www.generalist.com/p/the-token-budget-problem
—
Timestamps
(00:00) Intro
(03:50) Noether’s theorem explained
(06:45) How the search for what’s conserved informs Matan’s work
(10:53) Why there will always be more problems to solve
(11:58) The resource allocation problem of the AI era
(15:54) Factory’s mission: bringing autonomy to software engineering
(18:28) How Factory decides what to build next
(20:10) Why Factory abstracts away model choice
(22:07) How Factory wins enterprise customers
(23:15) Matan’s take on the SpaceX-Cursor deal
(27:48) Why open-weight models matter
(29:19) Anthropic’s Fable 5 release and the debate over AI guardrails
(35:33) How Matan got into string theory
(38:21) Working with Juan Maldacena
(41:53) Startup founders vs. theoretical physicists
(46:15) Rethinking physics and redefining his identity
(51:29) Discovering AI and code generation
(52:53) The origins of Factory
(55:52) Lessons from Factory’s first few years
(59:58) Learning to push back and finding the holes in his knowledge
(1:03:17) Factory’s culture and values
(1:08:11) Matan’s predictions for the future of AI and Factory
(1:10:49) Final meditations
—
Follow Matan Grinberg
LinkedIn: https://www.linkedin.com/in/matan-grinberg
X: https://x.com/matanSF
Website: https://factory.ai
—
Resources and episode mentions: https://www.generalist.com/p/the-token-budget-problem⁠
—
Production and marketing by penname.co. For inquiries about sponsoring the podcast, email jordan@penname.co.

Yash Patil is the 23-year-old founder and CEO of Applied Compute, a $1.3 billion company helping businesses train custom AI models on their own data: smaller, cheaper, and purpose-built for the work they actually do. Before founding the company, Yash dropped out of Stanford and spent two years at OpenAI working on post-training infrastructure and Codex. He left with one core conviction: every company that runs its critical workflows on someone else’s model is building on shifting sand. Applied Compute is his answer to that problem, already serving customers including DoorDash, Cognition, and Mercor.

In our conversation, we explore:
Why “own or be owned” is becoming existential for any company that relies on frontier AI models
What it was like inside OpenAI the weekend the board fired, and then reinstated, its CEO
Why post-training is where competitive advantage is now being built, and what reinforcement learning with verifiable rewards actually is
Why evals have become the new production environment, and why companies will never share them with frontier providers
How a specialized model built for DoorDash outperformed frontier models on a narrow, high-value task
Why cost, not capability, is now the primary driver pushing companies toward custom models
Why Yash believes AI’s transformation of the economy will unfold over decades, and why near-term fears about mass job displacement are misplaced
—
Thank you to the partners who make this possible
Brex: The intelligent finance platform.
Guru: The AI source of truth for work.
Persona: Trusted identity verification for any use case.
—
Transcript: https://www.generalist.com/p/own-or-be-owned-why-every-company
—
Timestamps
(00:00) Introduction
(03:50) Fable 5 and the case for owning your own models
(09:22) Why Applied Compute is betting on custom AI models
(12:30) Yash's early influences and first projects
(17:42) His brief time building at Stanford
(19:29) Leaving Stanford for OpenAI
(25:58) Inside OpenAI during Sam Altman's firing
(28:18) What Yash admires about Sam Altman
(29:43) Teaching models to reason
(35:39) The core insight behind Applied Compute
(39:40) How Applied Compute works with its customers
(45:55) Why model training never ends
(48:56) Why not every task needs a frontier model
(51:25) The culture and people of Applied Compute
(54:50) Applied Compute's training infrastructure
(58:43) The coming compute crunch and other predictions
(1:03:48) Final meditations
—
Follow Yash Patil
X: https://x.com/ypatil125
Website: https://yashpatil.me
LinkedIn: https://www.linkedin.com/in/yash-s-patil
—
Resources and episode mentions: https://www.generalist.com/p/own-or-be-owned-why-every-company⁠
—
Production and marketing by penname.co. For inquiries about sponsoring the podcast, email jordan@penname.co.

Bryon Hargis is the co-founder and CEO of Castelion, a defense startup building low-cost hypersonic missiles designed to be manufactured at scale. Before founding Castelion, Bryon spent more than a decade at Johns Hopkins Applied Physics Laboratory and nearly six years at SpaceX, where he worked on national security space programs and saw firsthand how iterative engineering and manufacturing speed could reshape aerospace. Castelion’s first missile, Blackbeard, is slated for integration on the Navy’s F/A-18 Super Hornet in roughly a year.
—
In our conversation, we explore:
Why Bryon believes building missiles is paradoxically essential to maintaining peace
The game theory behind warfare and why tit-for-tat strategies require credible middle-ground responses
How China’s 2021 hypersonic test revealed not just a capability gap but a manufacturing and cost advantage
Why traditional aerospace processes—optimized for low risk and high cost—can’t compete with rapid iteration
What Bryon learned in his first week at SpaceX (after 12 years in traditional aerospace)
Why building a carrier-based, air-launched hypersonic missile as a first product was the hard but right choice
How focusing on manufacturability and cost over maximum capability can produce more effective deterrence
Why the person who adapts faster in warfare always wins, and how that shapes Castelion’s philosophy
—
Thank you to the partners who make this possible
.tech domains: An identity for builders at their core.
Ahrefs Brand Radar: Find your brand in AI results.
Persona: Trusted identity verification for any use case.
—
Timestamps
(00:00) Intro
(04:01) Why America needs hypersonic missiles
(07:13) China’s edge in hypersonics
(12:05) The missing middle ground in deterrence
(18:05) Preventing warhead ambiguity
(19:40) How hypersonics differ from ballistic missiles
(25:05) The economics of defensive vs. offensive systems
(28:21) How SpaceX differs from traditional aerospace
(37:40) Why Bryon chose to build in defense over space
(42:42) Key factors that drove Castelion’s success
(48:28) Designing Blackbeard, Castelion’s first hypersonic missile
(1:01:06) The importance of lower costs and quicker manufacturing
(1:10:04) Book recommendations
—
Follow Bryon Hargis
LinkedIn: https://www.linkedin.com/in/hargsb
X: https://x.com/hargsb
—
Resources and episode mentions: https://www.generalist.com/p/what-america-is-missing-between-sanctions
—
Production and marketing by penname.co. For inquiries about sponsoring the podcast, email jordan@penname.co.

Davide Asnaghi is the co-founder and CEO of Diode, a Brooklyn-based startup using AI to design and manufacture circuit boards in the United States.

Before Diode, Davide worked on Apple’s Special Projects Group and spent time in Hong Kong and Shenzhen studying Asia’s electronics manufacturing ecosystem. That experience convinced him that PCB design, despite powering everything from smartphones and satellites to medical devices and autonomous systems, remained one of the most overlooked layers of the tech stack.

Since its founding just two years ago, Diode has landed Physical Intelligence and Saronic as customers and partnered with Anthropic to help Claude become a better electrical engineer. The company’s ultimate ambition: to make hardware as nimble as software.

In our conversation, we explore:
Why the West outsourced PCB manufacturing to Asia in the 2000s and why bringing it back matters for American competitiveness
What Shenzhen’s manufacturing culture does better than Silicon Valley (and what the U.S. can learn from it)
How Diode’s models can one-shot much of schematic design and compress hardware timelines from months to weeks
The three-week YC pivot that transformed Diode from a design validation tool into a full-stack manufacturer
Why circuit boards are the “forgotten middle child” between silicon and software
How Diode partners with Anthropic to make LLMs better electrical engineers
What it takes to build a hardware company in 2025—and why the talent bar must stay incredibly high
How Italian, American, and Chinese cultures shaped Davide’s approach to entrepreneurship and manufacturing
—
Thank you to the partners who make this possible
.tech domains: An identity for builders at their core.
Guru: The AI source of truth for work.
Brex: The intelligent finance platform.
—
Transcript: https://www.generalist.com/p/our-goal-is-to-build-an-electrical-engineer
—
Timestamps
(00:00) Intro
(04:15) Why Davide calls himself a copper merchant
(05:53) Diode’s mission to rebuild PCB manufacturing in the U.S.
(07:58) What success looks like
(09:00) Growing up in northern Italy and spending a year in Minnesota
(13:14) Why Italy produces fewer venture-backed founders
(15:30) Why Hong Kong accelerated Davide’s learning
(19:09) Silicon Valley vs. Shenzhen
(22:05) What Davide learned in Apple’s Special Projects Team
(24:11) Why Davide left Apple after two years
(26:54) Meeting his co-founder, Lenny
(29:32) How Davide uncovered the need for better PCB design and manufacturing
(33:23) PCB manufacturing in Asia, and Diode’s approach
(41:29) The YC pivot that changed Diode’s business
(44:39) Inside Diode’s customer journey
(48:10) Where the value is in electronics manufacturing, and Davide’s AGI thesis
(51:30) What separates a working board from a great one
(55:32) Where Diode fits in the electronics stack
(59:55) Diode’s early near-death moment and long-term vision
(1:02:30) Diode’s exceptionally high bar for hiring
(1:04:48) Where Davide gets his best ideas
(1:07:00) Final meditations
—
Follow Davide Asnaghi
LinkedIn: https://www.linkedin.com/in/d-asnaghi
X: https://x.com/davideasnaghi
GitHub: https://hexdae.github.io
—
Resources and episode mentions: https://www.generalist.com/p/our-goal-is-to-build-an-electrical-engineer⁠
—
Production and marketing by penname.co. For inquiries about sponsoring the podcast, email jordan@penname.co.