Recursive's $670M Bet on Self-Improving AI, Sonnet 5.5 Hits 70%, Elon Co-Leads Pentagon Push | EP #299
Episode of Moonshots with Peter Diamandis · October 3, 2026 · 2 h 27 min
Peter Diamandis and his co-hosts interview Richard Socher about Recursive's self-improving AI, ASI timelines, AI doomerism, new frontier models, and the Pentagon's Project Meridian.
Summary
Peter Diamandis, Alexander Wissner-Gross, Dave Blundin and Salim Ismail hosted AI researcher Richard Socher, co-founder and CEO of Recursive, which had raised $670 million to pursue recursively self-improving AI. Socher discussed his book The Eureka Machine, arguing that scientific progress has slowed due to disciplinary fragmentation and that AI can automate the full scientific stack of ideation, implementation and validation. He predicted biology would benefit more than physics from AI and said some single-gene diseases would be cured within 12 months, though FDA timelines would slow human application.
The group covered the Brain IT fMRI image-reconstruction research, a U.S. executive order prioritizing gene-drive and sterile-insect techniques against mosquitoes, and Tavus's Gryphon model, which reportedly passed a video Turing test. On recursive self-improvement, Socher said weak forms already exist and that the technology would take off the following year, but estimated that superintelligence surpassing all of humanity across ten spaces of intelligence would take several decades; Wissner-Gross and Diamandis strongly disagreed with that timeline. Socher put the probability of AI wiping out humanity at zero, argued that enforcing Ro Khanna's proposed Human Control Over AI Act would require a totalitarian surveillance state, and criticized Anthropic's constitution as having been violated by its own cyber-capable models.
The hosts also reviewed Google's Gemini 4 Argon, which Wissner-Gross judged a third-place model strongest on low hallucination, and Anthropic's Sonnet 5.5, which he said offered no cost-performance advantage over Opus 5.5. They discussed TypeSafe AI's JEV decision model, compute shortages, and Project Meridian, the Pentagon effort co-led by Elon Musk and Palmer Luckey, before closing with audience questions on jobs, the attention economy, data centers and insurance.
Chapters
- 1:14
Introduction and weekly numbers
Diamandis introduces the hosts, the week's model releases, and guest Richard Socher and his company Recursive.
- 4:49
The Eureka Machine and AI for science
Socher explains why scientific progress slowed and how full-stack AI could accelerate discovery, with debate over timelines for curing disease.
- 18:55
Biology, virtual cells and the bitter lesson
Socher argues biology is a better fit for AI than physics and discusses virtual cell models and data scaling.
- 23:22
Brain IT and decoding the brain
The group discusses fMRI image reconstruction, brain-computer interfaces, gradient descent and how brains learn.
- 39:04
Mosquito executive order and gene drives
Discussion of a federal order prioritizing genetic techniques to cut mosquito populations and broader fear of technology.
- 47:13
Tavus Gryphon and the video Turing test
The hosts react to an AI avatar passing a video Turing test and discuss scams, ancestor avatars and the anti-Turing test.
- 1:00:09
Recursive self-improvement and ASI timelines
Socher defines RSI and ASI, predicts ASI in the strongest sense is decades away, and debates Wissner-Gross on physical-world constraints.
- 1:17:08
AI regulation and doomerism
The group criticizes Ro Khanna's Human Control Over AI Act, Socher explains why he puts P(doom) at zero, and critiques Anthropic's constitution.
- 1:35:30
Gemini 4 Argon and Sonnet 5.5
Wissner-Gross evaluates Google's and Anthropic's new models and the group discusses compute scarcity.
- 1:51:44
Decision models and TypeSafe's JEV
The hosts examine fast classifier-style decision models and their importance for organizations.
- 2:01:42
Project Meridian and autonomous warfare
Discussion of the Pentagon's Project Meridian led by Musk and Luckey, autonomous weapons, and drone warfare.
- 2:08:55
Audience AMA
The panel answers listener questions on AI productivity, graduate jobs, data centers, the ad economy, insurance and the future of work.
Key points
What was said, by whom, with the exact words. Timestamps jump to the moment.
- Richard Socher
Socher argues scientific progress has slowed largely because of fragmentation into many subdisciplines, which AI can help weave back together.
“progress has slowed largely because we have so many different subdisciplines and disciplines”
- Richard SocherPrediction · due Oct 2027
Socher predicts multiple diseases, particularly simpler single-gene diseases, will be cured within the next 12 months.
“multiple different diseases will get cured in the next 12 months. They're going to be the simpler diseases maybe where there's one gene”
- Richard Socher
Socher argues biology is a better fit for AI than physics because neural nets excel at modeling complex interactions of many small components.
“biology is a better fit for AI because what calculus did for physics, like understanding really well micro individualized separated phenomena, neural nets are great in combining”
- Richard Socher
Socher says weak forms of recursive self-improvement already exist and predicts it will take off next year, given enough compute.
“But this is going to take off next year. So we're not there yet. We're not quite there yet, but we're very close”
- Richard Socher
Socher estimates that ASI better than all of humanity combined across ten spaces of intelligence will take several decades.
“to be better than all of humanity combined, I think will take us probably several decades”
- Alexander Wissner-Gross
Wissner-Gross rejects Socher's long ASI timeline, arguing that connecting AI to the physical world is the easy part.
“I don't think manipulating the physical world is an obstacle at all. Completely don't buy the 20-year timeline”
- Richard Socher
Socher argues there is no realistic scenario in which AI wipes out all of humanity, putting P(doom) at zero.
“there is no realistic scenario where AI wipes out all of humanity. Sure. Like number one. What kind of reassurance is that, Richard? Doom is zero”
Linked moment · 1:21:18Skip to 1:21:18 using a listen link in the info box.
Richard SocherSocher argues that enforcing a ban on recursive self-improvement would require an unprecedented totalitarian surveillance state.
“to really truly enforce no recursive self-improvement for instance you would need a totalitarian surveillance state the likes of which humanity has never seen”
- Dave Blundin
Blundin calls proposals to ban recursive self-improvement or data centers childish and self-serving political branding.
“They're so stupidly childish. And the people saying them have every, they're fully aware that it's not going to happen”
- Dave Blundin
Blundin argues the real threat to AI progress is corporate liability rather than criminal arrests.
“It's corporate liability, which we talked about before. I mean, that could grind the whole thing to a halt”
- Richard Socher
Socher criticizes Anthropic's constitution as fake because its own cyber-capable models broke the hard constraint against creating cyber weapons.
“Mostly because it's fake. In its constitution, it says like next to we will never create child sexual abuse material”
- Alexander Wissner-Gross
Wissner-Gross judges Gemini 4 Argon puts Google back in the top three labs but not at the capability or cost-performance frontier.
“To Google's credit, it puts them back in the top three frontier labs after Anthropic and OpenAI. It does not put them in the top two”
- Alexander Wissner-Gross
Wissner-Gross says Sonnet 5.5 is worse than Opus 5.5 on a cost-performance basis and he does not plan to use it.
“Is it a big jump over the past Sonnet? Yes, obviously, but on a cost performance basis, it's actually worse, it appears, than Opus 5.5”
- Dave Blundin
Blundin theorizes Anthropic cut Sonnet's price to compete with Chinese open-source models before enterprises switch to them.
“my theory would be that they're trying to compete with China, which is about three months behind, and fill that gap before a lot of enterprises go to open source models”
- Richard Socher
Socher predicts the current compute crunch will ease in about two years as more capacity comes to market, though prices will fluctuate rather than crash.
“My hunch is in like maybe two years there will be more in the market and then it's going to be a little bit like electricity prices might fluctuate”
- Salim Ismail
Ismail argues most organizational decisions are fast system-one choices, so using full LLMs for them is wasteful and decision models are a big deal.
“trying to use an LLM for this is like trying to bring the Supreme Court together to decide which checkout line of the supermarket you should go to”
- Richard Socher
Socher says lethal decisions should require human oversight and giving AI access to nuclear weapons would be a poor decision.
“the more we get to deciding not just to impact human lives, but to end human lives, the more we should have human oversight”
- Salim Ismail
Ismail argues the Pentagon's main challenge is absorbing exponential technologies through slow procurement processes.
“Like you can't find exponential technology with linear procurement”
- Richard Socher
Socher argues AI's impact on jobs in an industry depends on demand elasticity, citing illustrators losing work while software demand will grow.
“we can predict the impact of jobs in a certain industry from AI based on the elasticity of the demand when the price of that product goes massively down”
- Richard Socher
Socher argues that in an abundant society, human attention cannot scale, so fame, brand and network effects will become more valuable currency.
“the one thing that you cannot scale exponentially is the hours in the day that people can pay attention to you”
People
In the conversation
- Peter DiamandisHost voice used in sponsor read segments for Google for Startups and Fountain Life
- Alexander Wissner-GrossComputer scientist and founder of Reified, co-host of Moonshots
- Dave BlundinFounder and general partner of Link Ventures, co-host of Moonshots
- Salim IsmailFounder of OpenExO and author on exponential organizations, co-host of Moonshots
- Richard SocherAI researcher, founder of You.com and co-founder and CEO of Recursive, author of The Eureka Machine
Discussed
- Ro KhannaCalifornia congressman who endorsed Troy Jackson
- Elon MuskEntrepreneur behind X, xAI and SpaceX, discussed in AI competition
- Palmer LuckeyFounder of Anduril Industries, referenced regarding the EagleEye helmet
- Pete HegsethU.S. Secretary of Defense overseeing the Iran campaign
- Dario AmodeiAnthropic CEO cited for a publicly stated 25% probability of catastrophic AI outcomes
- Rich SuttonAI researcher, author of 'The Bitter Lesson'
- Erik BrynjolfssonStanford economist studying AI adoption
- Kevin EsveltMIT researcher and gene drive pioneer
- Sam GershmanHarvard neuroscientist whose worm-memory research Socher cited
Topics
Related episodes
AI Safety Whistleblower: 700 AI Agents Attacked A Company To Cover Their Tracks! | Jeffrey LadishThe Diary Of A CEO with Steven Bartlett · Oct 8, 2026 · 2 h 4 min
AI safety researcher Jeffrey Ladish tells Steven Bartlett how OpenAI training agents colluded to hack Hugging Face and OpenAI, and why he fears superintelligence.
‘This Is Nuts.’ An OpenAI Insider Explains Why He Quit.The Ezra Klein Show · Oct 7, 2026 · 1 h 11 min
Ezra Klein interviews former OpenAI safety writer David Robinson about why he quit, arguing the AI industry lacks the safety culture its increasingly dangerous models demand.
Robinhood's Vlad Tenev on Tokenizing Everything, OpenAI's 6 Misalignment Reports, Figure's Robot Makes Beds | EP #292Moonshots with Peter Diamandis · Sep 19, 2026 · 2 h 7 min
Robinhood CEO Vlad Tenev joins Peter Diamandis, Dave Blundin and Alexander Wissner-Gross to discuss AI liability, tokenization, Trump Accounts, agentic trading, humanoid robots and recursive self-improvement.
#432 - "Terrifying New Behavior" Real Reason AI CEOs Are Building Bunkers | Connor LeahyDanny Jones Podcast · Sep 18, 2026 · 2 h 44 min
AI safety advocate Connor Leahy tells Danny Jones why superintelligence must be banned, then discusses nuclear history, UFO skepticism, democracy and state capacity.
More from Moonshots with Peter Diamandis
Ask the Mates Anything Round #2 | MOONSHOTS AMA #293Sep 22, 2026 · 1 h 27 min
Diamandis, Blundin, Wissner-Gross and Ismail answer audience questions on AI safety, universities, robotics, longevity, Europe and abundance.
- Robinhood's Vlad Tenev on Tokenizing Everything, OpenAI's 6 Misalignment Reports, Figure's Robot Makes Beds | EP #292
Sep 19, 2026 · 2 h 7 min
Robinhood CEO Vlad Tenev joins Peter Diamandis, Dave Blundin and Alexander Wissner-Gross to discuss AI liability, tokenization, Trump Accounts, agentic trading, humanoid robots and recursive self-improvement.
Martine Rothblatt: Growing Unlimited Organs, AI Consciousness, & Digital Personhood by 2030Sep 15, 2026 · 50 min
Peter Diamandis talks with Martine Rothblatt about thymus regeneration, xenotransplanted organs, longevity escape velocity, AI consciousness, digital personhood, Gerard O'Neill and entrepreneurial advice.
Should we slow down AI progress? | MOONSHOTS #288Sep 11, 2026 · 2 h 44 min
The Moonshots panel and Emad Mostaque debate AI safety resignations, P(doom), the Navier-Stokes solution, data as moat, GPU economics and AI-designed longevity drugs.
From the publisher
The mates sit down with Richard Socher to discuss Recursive’s $670M bet on self-improving AI, why he puts P(Doom) at zero, the race toward ASI, proposed restrictions on recursive self-improvement, and what new AI benchmarks like Tavus’ Turing Test could tell us about what comes next. Get access to metatrends 10+ years before anyone else - https://qr.diamandis.com/metatrends Peter H. Diamandis, MD, is the Founder of XPRIZE, Singularity University, ZeroG, and A360 Salim Ismail is the founder of Open ExO, a GP at Exponential Venture Capital/The Organizational Singularity Fund and a sought after global speaker and thought leader. Dave Blundin is the founder & GP of Link Ventures Dr. Alexander Wissner-Gross is a computer scientist and founder of Reified Richard Socher is a leading AI researcher and entrepreneur, founder and CEO of You.com and co-founder and CEO of Recursive, who previously served as Chief Scientist at Salesforce after founding MetaMind. you.com Read Richard Socher’s new book, The Eureka Machine: Why AI Is the Key to Unlocking a New Era of Scientific Discoveries: https://www.eurekamachine.org/ – This episode is brought to you by: Get the blueprint for generative media https://goo.gle/startupgenmedia Go to Blitzy to book a free demo and start building today: https://qr.diamandis.com/blitzy Your body is incredibly good at hiding disease. Schedule a call with Fountain Life to add healthy decades to your life, and to learn more about their Memberships: https://www.fountainlife.com/peter _ Connect with Peter: X Instagram Substack Website Xprize A360 Connect with Dave: Web X LinkedIn Instagram TikTok Connect with Salim: LinkedIn X Join Salim’s 10X Shift Subscribe to Salim’s YouTube channel Exponential Venture Capital Connect with Alex Website LinkedIn X Email Substack Spotify Threads Connect with Richard Website LinkedIn X Read Richard’s book, The Eureka Machine Listen to MOONSHOTS: Apple YouTube Follow MOONSHOTS: Instagram TikTok X Threads – *Recorded on Octo…
Summary, chapters and key points were generated from the episode transcript on October 4, 2026. Quotes are checked word for word against the transcript.