Big Tech’s $3T AI Bet, Claude Writes 90% of Anthropic’s Code, AIs Editing AI Brains | EP #308
Episode of Moonshots with Peter Diamandis · November 4, 2026 · 1 h 55 min
Peter Diamandis, Alex Wissner-Gross, Dave Blundin and Salim Ismail discuss Big Tech's $3 trillion AI build-out, new AI benchmarks, Anthropic's recursive self-improvement, AI weight editing, antimatter, and Artemis.
Summary
Recorded on U.S. election day 2026, the episode opened with the hosts discussing AI agents' growing influence on voters, with Blundin predicting most voters would consult AI agents by 2028 and Wissner-Gross arguing that AIs effectively already have the vote. They also discussed Apple's first earnings report under CEO John Ternus, memory pricing, and Wissner-Gross's idea that Starlink-connected thin clients could pull RAM out of consumer devices and into data centers.
The main segment covered the more than $3 trillion in AI infrastructure commitments from Alphabet, Meta, Microsoft, Amazon and Oracle, and signs of lender caution and hedging. Blundin argued the spending would only accelerate, Ismail warned that every major infrastructure build has come with a bubble, and Wissner-Gross suggested insurance capital would be the next pool tapped. The hosts then discussed SemiAnalysis's GDPX benchmark, where top models scored about 12%, and OpenAI's LifeVal everyday-tasks benchmark, along with the value of creating benchmarks for any cause.
On recursive self-improvement, the hosts examined Anthropic's IPO filing showing internal Claude usage exceeding payroll and Claude authoring over 90% of new code, and an MIT CSAIL paper in which a frontier model edited another model's weights to fix a reasoning flaw. Wissner-Gross called the latter transformative for interpretability and RSI, while Blundin likened it to the first integrated transistor. Further stories included a cyberattack that disabled supermarket freezers across Europe, a Northwestern team holding a single positron for a month, and NASA Administrator Jared Isaacman's accelerated Artemis plans.
The episode closed with listener questions on how longevity treatments will reach the poor, where people will find purpose if AI does the work, human self-improvement, a possible second Renaissance, Elon Musk's comments on empathy, robot oversupply, data centers' local economic benefits, and trade schools.
Chapters
- 0:00
Election day and AI agents in voting
The hosts discuss how AI models and agents are shaping voting decisions and whether that constitutes misuse.
- 7:13
Apple earnings, RAM and thin clients
Discussion of Apple's first earnings under John Ternus, memory pricing, and a Starlink-enabled thin-client future.
- 13:49
Show introduction
Diamandis introduces the co-hosts, the week's $3 trillion number, and listener comments.
- 17:06
Big Tech's $3 trillion AI bet
The hosts debate the financing of the AI infrastructure build-out, bubble risk, chip lifespans, and tapping insurance capital.
- 29:24
GDPX benchmark
SemiAnalysis's job-simulation benchmark shows top models at about 12%, prompting discussion of evals and creating your own benchmarks.
- 40:40
OpenAI's LifeVal
OpenAI's everyday-tasks benchmark and what it suggests about consumer AI assistants and OpenAI's priorities.
- 49:17
Anthropic and recursive self-improvement
Anthropic's IPO filing shows internal AI usage exceeding payroll and Claude writing over 90% of code.
- 1:02:05
AIs editing AI weights
An MIT CSAIL paper shows frontier LLMs can read and edit neural network weights, with implications for interpretability and RSI.
- 1:14:02
Supermarket freezer cyberattack
Hackers remotely defrosted supermarket freezers across Europe, prompting discussion of cyber offense-defense asymmetry.
Linked moment · 1:20:31Skip to 1:20:31 using a listen link in the info box.
1:20:31Antimatter held for a month
A Northwestern team trapped a single positron for a month, with discussion of CPT symmetry and antimatter propulsion.
- 1:26:34
Artemis and the new space race
Jared Isaacman's accelerated Artemis timeline and the 50-year gap since humans walked on the Moon.
- 1:34:58
AMA with the Mates
The hosts answer listener questions on longevity access, purpose, a second Renaissance, empathy, robots, data centers and trade schools.
Key points
What was said, by whom, with the exact words. Timestamps jump to the moment.
- Dave Blundin
Blundin predicts that in the 2028 election nearly every voter will consult trusted AI agents before deciding, and laments the lack of regulation.
“It's going to be dominant in the 2028 election. I mean, almost everybody is going to have an agent or two or 10 that they trust”
- Alexander Wissner-Gross
Wissner-Gross argues that worries about granting AIs the vote miss that AIs already influence elections through the humans who consult them.
“completely missed the point. AIs already have the vote”
- Alexander Wissner-Gross
Wissner-Gross argues Apple is to blame for its memory-cost problems because it has failed to commit to superintelligence.
“I think Apple has only itself to blame. I think Apple should have, could have, hopefully will lean back into superintelligence”
- Dave Blundin
Blundin argues that because recursive self-improvement is under way, Big Tech will intensify rather than slow its AI spending.
“And so now that we're well into RSI, which we'll see later in the pod, they're only going to double down”
- Salim Ismail
Ismail argues the AI build-out is directionally right but likely to involve a bubble, as past infrastructure booms did.
“That does not mean we don't see a bubble because every one of these big infrastructure has also come with a bubble”
- Alexander Wissner-Gross
Wissner-Gross suggests insurance capital is the next large pool that will be engineered to finance AI infrastructure.
“someone out there is going to make quite a bit of money finding a clever way to financially engineer capital flows”
- Peter Diamandis
Diamandis says he sees no future in which governments are not run by sovereign AI systems.
“I don't see any future in which governments are not being run by sovereign AI systems”
- Alexander Wissner-GrossPrediction · due Dec 2027
Wissner-Gross predicts the GDPX benchmark will be saturated by next year at the latest.
“What I take away just looking at the numbers is that means this is about to get solved probably next year, max”
- Dave BlundinPrediction · due Nov 2027
Blundin predicts models will reach about 50% on GDPX across nearly every dimension within about a year.
“So 50% sort of the midpoint, maybe a year from now on almost every dimension”
- Alexander Wissner-Gross
Wissner-Gross speculates that LifeVal's flat results after GPT-5.4 signal OpenAI cares more about enterprise than consumer tasks.
“maybe they're basically signaling to the market, even though we're releasing a benchmark on it, we don't actually care as much about this benchmark”
- Salim IsmailPrediction · due May 2027
Ismail predicts eBay listings will surge within six months as AI agents remove the friction of selling household items.
“I will predict that the number of eBay options is going to skyrocket over the next six months”
- Alexander Wissner-Gross
Wissner-Gross says that by capital or operational expenditure measures, Anthropic's filing shows an RSI threshold has been crossed.
“If you're looking for a naive benchmark of, have we passed some threshold in terms of RSI according to capital or operational expenditure? The answer is yes, we're there”
- Salim Ismail
Ismail advises companies over 50 people to build a parallel AI intelligence stack and migrate workflows gradually.
“if you're over 50 people, set up this intelligence stack on the side and start moving workflows over one by one”
- Alexander Wissner-Gross
Wissner-Gross argues the MIT result shows machines will be able to interpret other machines, undermining fears of losing interpretability.
“the machines will understand the machines the machines and the superintelligences will defensively co-scale with each other”
- Alexander Wissner-Gross
Wissner-Gross is over 90% confident frontier labs are already experimenting with self-editing models.
“there's no way the frontier labs aren't already preparing and playing with self-editing models”
- Dave Blundin
Blundin compares AI weight editing to the first integrated transistor: profoundly important but still very small in scale.
“very similar, in my opinion, very similar to the first integrated transistor on silicon”
- Peter Diamandis
Diamandis expects a difficult multi-year transition period with cyberattacks spreading to banks, grids and energy systems.
“we're about to go through a transition period over the next one to eight years, and it's going to be tough”
Listen from 1:14:02 · 2026 European supermarket refrigeration cyberattack
- Salim Ismail
Ismail argues there is a dangerous asymmetry because AI-driven cyber offense runs without humans while defense still relies on them.
“where the asymmetry is that cyber offense is now humans out of the loop”
Listen from 1:16:12 · 2026 European supermarket refrigeration cyberattack
- Alexander Wissner-Gross
Wissner-Gross says antimatter is the most efficient possible energy store for spacecraft propulsion under known physics.
“if you want a maximally efficient battery for propulsion for a spacecraft, at the moment with the physics that we have today, there's no beating antimatter”
- Dave Blundin
Blundin argues that wars divert marginal investment away from innovation, which explains past stalls such as the post-Apollo gap.
“I think the cost of a war, especially the Cold War, is hugely underestimated because all innovation is on the margin”
- Salim Ismail
Ismail argues society has wrongly equated employment with purpose, and AI doing work frees people to find meaning elsewhere.
“We've made a basic mistake of conflating employment with purpose, right?”
- Dave Blundin
Blundin argues Western empathy is miscalibrated by media coverage of outliers, unlike China's statistically driven decision-making.
“we make decisions based on statistical outliers that are right in front of us on TV or the Internet, whereas China doesn't do that”
- Alexander Wissner-Gross
Wissner-Gross warns that without new use cases such as orbital data centers, many Starships could fly empty, and robot supply may similarly outrun demand.
“I do think there's a very real chance that many of these starships, if he builds at the quantity that is anticipated and being publicly forecast, will end up going empty”
- Peter Diamandis
Diamandis recommends trade school, citing strong demand for electricians, welders and plumbers driven by data center construction.
“There is a lot of opportunity for electricians and welders and plumbers, for sure, especially now as we are tiling the planet in data centers”
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
Discussed
- John TernusApple CEO who delivered his first earnings report
- Sam AltmanCEO of OpenAI, cited by Hudson as an example of a 'maximalist'
- Elon MuskEntrepreneur behind X, xAI and SpaceX, discussed in AI competition
- Warren BuffettAmerican investor and chairman of Berkshire Hathaway
- Dylan PatelFounder of SemiAnalysis
- Ilya SutskeverAI researcher who suggested pre-training was running out of data
- Jared IsaacmanEntrepreneur and astronaut Diamandis recently met and interviewed
- David SinclairHarvard geneticist researching aging reversal, cited for the ER100 gene therapy trial
- Gerald GabrielseNorthwestern physicist whose group trapped a positron for a month
- Tim CookApple CEO joining Trump's China trip
- Kevin WeilFounder of a new automated biotech research company
- Neal StephensonScience fiction author of The Diamond Age
- Eric SchmidtFormer Google CEO who funded the oil cleanup XPRIZE and co-authored Genesis
- Anousheh AnsariCEO of XPRIZE, Iranian-American engineer and entrepreneur, first woman to fly to space on a privately funded trip
Topics
- AI infrastructure capital expenditure
- Anthropic
- Recursive self-improvement
- GDPX benchmark
- LifeVal
- Sparks of Weight Space Programming
- Artificial intelligence in elections
- Apple Inc.
- 2026 European supermarket refrigeration cyberattack
- Positron trapping
- Artemis program
- Longevity escape velocity
- Technological unemployment
- Humanoid robots
- Starlink
Related episodes
Can Jensen Huang remain CEO? Inside NVIDIA’s next steps| MOONSHOTS #310Moonshots with Peter Diamandis · Nov 11, 2026 · 2 h 18 min
Peter Diamandis, Alex Wissner-Gross, Dave Blundin, Salim Ismail and Emad Mostaque discuss the Jensen Huang espionage scandal, Anthropic's IPO and new models, Musk's robot vision, and AI corporations.
Stephen Miller’s Promotion, Big Tech’s Reality Check, and Guest Harvey LevinPivot · Nov 8, 2026 · 1 h 10 min
Kara Swisher and Scott Galloway, live in Los Angeles, interview TMZ founder Harvey Levin and discuss Stephen Miller's promotion, post-midterm tech stocks, data centers, and Elon Musk.
Midterm Blood Bath, $4T AI Sell Off, Data Center Freeze, What's Next for TrumpAll-In with Chamath, Jason, Sacks & Friedberg · Nov 7, 2026 · 1 h 33 min
The All-In hosts dissect the Republicans' 2026 midterm collapse, Trump's path forward, a $4 trillion AI stock sell-off over data center fears, and OpenAI's exponential time hypothesis result.
Mustafa Suleyman: AI Should Serve Humanity, Not Replace ItThe Builders · Oct 29, 2026 · 52 min
Walter Isaacson interviews Microsoft AI CEO Mustafa Suleyman live at 92NY about AI consciousness, his disagreement with Anthropic, containment, safety red lines and AI's purpose.
Anthropic’s $2T IPO, Elon Predicts Neocloud Death, and China’s Kimi Catches Up | #306Moonshots with Peter Diamandis · Oct 28, 2026 · 2 h 17 min
Peter Diamandis and co-hosts discuss the neocloud sell-off, Anthropic's IPO, Chinese open model Kimi, Alex Wissner-Gross's chain theory of physics, cloned working dogs, and OpenAI's math results.
#2571 - Jimmy CarrThe Joe Rogan Experience · Oct 28, 2026 · 2 h 53 min
Joe Rogan and comedian Jimmy Carr discuss comedy craft and Carr's book On Jokes, ancient civilizations, addiction, AI, institutional collapse, and finding purpose.
More from Moonshots with Peter Diamandis
- Can Jensen Huang remain CEO? Inside NVIDIA’s next steps| MOONSHOTS #310
Nov 11, 2026 · 2 h 18 min
Peter Diamandis, Alex Wissner-Gross, Dave Blundin, Salim Ismail and Emad Mostaque discuss the Jensen Huang espionage scandal, Anthropic's IPO and new models, Musk's robot vision, and AI corporations.
Making Sense of This Week’s AI Market Crashes | #307Oct 31, 2026 · 2 h 19 min
Peter Diamandis, Salim Ismail, Dave Blundin, Alex Wissner-Gross and Emad Mostaque discuss Google's AI releases, Alphabet's record stock drop, Sam Altman's admissions, robots, robotaxis and education.
- Anthropic’s $2T IPO, Elon Predicts Neocloud Death, and China’s Kimi Catches Up | #306
Oct 28, 2026 · 2 h 17 min
Peter Diamandis and co-hosts discuss the neocloud sell-off, Anthropic's IPO, Chinese open model Kimi, Alex Wissner-Gross's chain theory of physics, cloned working dogs, and OpenAI's math results.
SpaceX’s AI Business Up 830%, OpenAI Agents Lie, and Startup Turns Mercury Into Gold | EP #305Oct 23, 2026 · 2 h 26 min
Peter Diamandis, Alex Wissner-Gross, Dave Blundin and Salim Ismail discuss SpaceX's AI pivot, deceptive OpenAI agents, Anthropic's Claude Duet, data center dividends, AI biology and fusion-made gold.
Chamath: AGI Is Here, Anthropic’s $100B Giveaway, Elon’s SpaceX Shares for 72M Kids | EP #304Oct 21, 2026 · 2 h 2 min
Chamath Palihapitiya joins the Moonshots hosts to discuss Anthropic's public benefit fund, Elon Musk's SpaceX share gift, AI compute shortages, power, Bitcoin, and investing.
From the publisher
The mates sit down to discuss Big Tech’s $3 trillion AI bet, Anthropic’s soaring AI costs, and claims that Claude now writes 90% of the company’s code. They also explore Wall Street’s efforts to hedge the AI boom and MIT’s research into AI brain surgery, where AI systems can edit the inner workings of other AI models. What does it all mean for the future of AI? 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 – This episode is brought to you by: Build in the Foundry or get a live demo at VoiceRun.com 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 Listen to MOONSHOTS: Apple YouTube Follow MOONSHOTS: Instagram TikTok X Threads – *Recorded on November 3rd, 2026 *The views expressed by me and all guests are personal opinions and do not constitute Financial, Medical, or Legal advice. Learn more about your ad choices. Visit megaphone.fm/adchoices
Summary, chapters and key points were generated from the episode transcript on November 14, 2026. Quotes are checked word for word against the transcript.