Subscribe to stay ahead of technology trends. Never miss future editions.
You.com founder Richard Socher builds self-improving superintelligence
For the past few years, the AI race has been about bigger models trained on more internet data. Our latest guest thinks that playbook is running out of road. As he puts it, a lot of people think the future is “kind of this prediction exercise, but it’s often a construction exercise.”
Meet Richard Socher, co-founder and CEO of Recursive and founder of You.com. He joined us to explain why AI’s biggest opportunity is speeding up scientific discovery, why some rare genetic diseases could be curable within the next 12 to 24 months, and why aging will be medicine’s “ultimate final enemy.”
You might not know this, but Richard was one of the researchers behind ImageNet and co-authored GloVe, two building blocks of modern AI. His first startup, MetaMind, was acquired by Salesforce, where he went on to serve as Chief Scientist. He then founded You.com, which reached a $1.5 billion valuation in 2025, and co-founded AIX Ventures. His new book, The Eureka Machine, argues that AI is the key to a new era of scientific breakthroughs. He wrote it over two and a half years of nights and weekends.
Recursive came out of stealth in May with $650 million raised at a $4.65 billion valuation. The round was led by GV and Greycroft, and Nvidia. Richard leads a team of seven co-founders who previously worked at Meta FAIR, Google DeepMind and OpenAI.
In our episode, we learn:
Why biology is AI’s next frontier. Why single-gene diseases are close to fixable, why biology still has to learn the “bitter lesson” that language AI already learned, and why the path runs through a virtual cell, then a virtual tissue, and eventually a full digital twin of a person
Why curing aging will take decades, not months. Every organ ages at a different speed, and clinical trials take years. The bar is far higher for drugs given to healthy people. Richard also explains why a longevity drug could create a multi-trillion-dollar company, and why scaling eventually makes every technology more equal
Why he isn’t buying calls to pause AI. He sees a “horseshoe theory” of AI optimism and doomerism, and explains why travel, farming and luxury handbags won’t change much with a thousand-times cheaper intelligence. He also covers why owning equity, and learning entrepreneurship, decides who wins from AI
Where startups can still win. Network effects, brand, data moats and what he calls “the moat of boring.” He explains why AI is turning the US into China’s no-SaaS market, and why China’s lead in open source works as soft power, much like Hollywood
Inside Recursive. How the team built Flint, a coding model that runs on a laptop, from scratch in three and a half months using its own “Eureka machine.” He also covers why eight co-founders still need one CEO, how he runs three organizations on 80-to-100-hour weeks, and why the West needs “better marketing for the future”
Watch or listen now across YouTube, Apple Podcasts, Spotify, and X
Download the transcript 👇
Timestamps
(0:00) Meet Richard Socher
(2:02) Richard’s new book
(5:17) Who solves humanity’s biggest problems first?
(8:12) How close are we to curing disease?
(11:30) How Richard would solve aging
(15:57) The opportunity in new drugs
(19:04) Is AI moving too fast?
(21:36) The superintelligence timeline
(25:40) How we all get wealthier
(28:45) Advice for hourly workers
(33:00) The biggest superintelligence opportunities
(35:32) Was Muse essential for Meta?
(36:20) Will big tech wipe out startups?
(40:05) US vs China: who wins?
(43:16) Meet Flint, the new coding agent
(46:10) Should founders just launch?
(47:23) Raising $500M+ and managing multiple co-founders
(49:29) How Richard manages his time
(51:42) Should founders focus on one company?
(54:40) Superintelligence in the next 12 months
Inside Recursive, Richard Socher’s $650M Bet on AI That Improves Itself
Some Rare Diseases Could Be Curable Within Two Years. Aging Will Take Decades.
Recursive is building self-improving superintelligence: AI that can come up with, test and ship its own improvements. It was founded by Richard Socher, who worked on ImageNet, co-authored GloVe, sold his first startup MetaMind to Salesforce, and then became Salesforce’s Chief Scientist. He went on to found You.com, now valued at $1.5 billion, and co-founded AIX Ventures. His new book, The Eureka Machine, argues that AI is the key to a new era of scientific discovery.
Recursive came out of stealth in May with $650 million raised at a $4.65 billion valuation. GV and Greycroft led the round, with Nvidia. Richard leads a team of eight co-founders who previously worked at Meta FAIR, Google DeepMind and OpenAI.
Why He Wrote The Eureka Machine
Richard wrote the book over two and a half years of nights and weekends. Part of the reason was frustration with where AI research had settled.
“It felt to me like the field of AI was a little bit stuck in just saying, ‘Oh, we’re just going to take this and use more internet data and scale up these existing models,’ when I think the time was ripe to actually expand the use of AI into the sciences.”
The field moved fast while he wrote. He had to rewrite chapters that said “someone should build a virtual cell” once major research groups started building one.
He also pushes back on the idea that AI mostly destroys jobs. We can lament the people who sold candles, he says, but electricity created more jobs than the candle industry ever had.
Rare Diseases First, Aging Last
Richard thinks diseases caused by a single faulty gene will be curable within 12 to 24 months. Each one is rare, but together they affect a lot of people.
“This isn’t just sci-fi anymore. This will happen.”
Babies born with genetic diseases have already been cured. Aging, he says, will be medicine’s “ultimate final enemy.” The reason is data. Large language models learned language from text that billions of people wrote online. Nothing like that exists for biology yet. Richard’s path forward runs from a virtual cell to a virtual tissue, and eventually to a full digital twin of a person.
“A lot of people think the future is kind of this prediction exercise, but it’s often a construction exercise. We need to actually build these things.”
If he weren’t running companies, he says, he would work on automating biological data collection.
Why Longevity Will Take Decades
Even with a perfect molecule, aging isn’t one problem. Every organ ages at its own speed. The bar for approval is also much higher for drugs given to healthy people. A late-stage cancer patient will accept long-term risks, but a healthy 40-year-old taking a longevity drug won’t.
Richard thinks whoever cracks longevity would build “a multi-trillion-dollar company.” He isn’t worried about the drug starting out expensive. Patents run out and generics follow, just as brick phones gave way to an iPhone that billionaires and teenagers both carry.
“What are other goods and services that currently only billionaires have access to? It’s personalized healthcare advice, it’s a personal assistant, it’s a personal tutor for your kids… they’re bottlenecked on intelligence.”
Why He Isn’t Buying the Pause
When we asked about recent calls from AI leaders to slow down, Richard was blunt.
“I think if the agents at my company were committing a lot of felonies, I would probably say, ‘Hey guys, we, as in my company, are moving too quickly.’”
To him, a global pause is unrealistic, and suspicious coming from labs that kept releasing new models the next week. It looks more like a way for those labs to “keep pole position.”
He proposes a “horseshoe theory” of AI: extreme optimism and extreme doomerism end up closer to each other than they look. In his view, many industries won’t change much even with a thousand-times cheaper intelligence:
Travel: people still want to see the pyramids in person.
Farming: a cow won’t make a thousand times more milk, and wheat won’t grow a thousand times faster.
Luxury: a handbag’s value as a status symbol won’t change.
Own the Output
Richard’s dividing line is equity.
“If you’re just getting paid by the hour and you have no equity in making things faster, better, cheaper, then AI is eventually probably going to come after whatever data you’re producing, and your job.”
His advice is to learn entrepreneurship. Look at the bottlenecks in your industry that exist because intelligence is expensive, and build for the moment that changes. He points to Germany’s Volkshochschule, adult-education colleges where people learn new skills mid-career, and to the pride in craft he sees in Japan. He believes handmade goods, live sports and entertainment will hold their value.
US vs. China
“It’s very obvious that the US is better at coming up with truly novel ideas and really pushing the frontier.”
But AI has reached the stage where scaling matters most, and China excels at making things faster, better and cheaper. Richard expects robot brains to come from Silicon Valley and robot bodies from China. He sees China’s lead in open-source AI as soft power, much like Hollywood, and says Recursive plans to compete there.
Flint in Three and a Half Months
Flint is a coding model that runs on a laptop, such as a high-end MacBook Pro. The team built it from scratch in about three and a half months. Its goal is to make a local model good enough to use every day.
“Because AI is code and now AI can code, this is also a really important milestone for us towards full recursive self-improving superintelligence.”
Recursive’s own “Eureka machine” handled the generating, building and testing of ideas. The researchers worked with that system rather than doing every step by hand.
The Moat of Boring
Network effects, brand and data moats still protect startups. Richard adds one more.
“People at Anthropic are not that excited about automating taxes. It doesn’t feel like, ‘Oh, we got one step closer to AGI because we automated taxes.’”
He also sees China as a preview of software’s future. Labor there was so cheap that companies hired people to rebuild expensive software instead of buying it, so China never developed a deep SaaS market. Now AI makes rebuilding cheap in the US too.
Eight Co-Founders, One CEO
How do eight co-founders agree on anything? Their titles don’t come with equal votes.
“A company could have a thousand co-founders, but 999 of them could just be individual contributors.”
Recursive’s $650M round came together in a few weeks. Richard runs three organizations on 80-to-100-hour weeks. He warns that you earn the right to do several things at once only when none of them is a “dumpster fire,” and that right “can disappear very quickly.”
What’s Next: Cures, Assistants and Better Marketing for the Future
Over the next year, Socher expects:
The first single-gene diseases cured.
Personal assistants for more people, something only millionaires had until recently.
One-time-use apps: his sales team now builds custom dashboard websites for a single account, something nobody would have spent time on before.
More self-driving as Waymo, Tesla and others expand to more cities.
More fear of AI, because dystopian stories get clicks.
His counterexample is China, where soap opera heartthrobs play engineers, pop stars dance with robots and cities put up billboards of famous scientists.
“They’re doing better marketing for the future, and I think that’s something we need more of in the West too.”
Where to connect with us
Follow Ollie on X: https://x.com/ollieforsyth
Follow Richard on X: https://x.com/RichardSocher
Visit You.com: https://you.com
Visit Recursive: https://recursive.com
Previous episodes include
See all previous episodes here 👉
Our partners:
Prelint: Keeps every decision reviewed, recorded, and traceable. Head over to Prelint to learn more.
VCX: The public ticker for private tech. Head over to GetVCX.com to learn more.
