ALL-IN · EXTRACTED
Can the AI Industry Regulate Itself? Stripe Wants PayPal, China Catches Up, NY Bans Datacenters
AI regulation, the PayPal acquisition wave, and why enterprise token spend is about to blow up CFO budgets. Three operational reads on where the money and power are actually moving.
With David Sacks, Chamath Palihapitiya, David Friedberg
Preview · 1 of 12 tactics free
"The government does not have the expertise to evaluate AI models. The criteria are changing too rapidly. You're going to very rapidly end up with a queue where all the models would be waiting to get tested." — David Sacks
This episode of the All-In Podcast brings together Jason Calacanis, David Sacks, Chamath Palihapitiya, and David Friedberg to work through three converging stories: the Demis Hassabis SRO proposal for AI regulation, the Stripe-Block bid for PayPal, and a wave of enterprise AI cost surprises that Ramp's data just made impossible to ignore. The pop read on AI regulation is that the industry is finally coming to its senses and agreeing to oversight. The actual operating system underneath is sharper: every concession to government becomes the floor for the next demand, and the five conditions that make an SRO workable are not in the current proposal. On payments, the story looks like a bold acquisition but the hosts argue it is the opening move in a direct assault on Visa and Mastercard rails. And on enterprise AI spend, a single Ramp stat, 21x token growth in one year, reframes the whole cost conversation from pennies to earnings misses. This playbook pulls the three highest-leverage reads from the conversation.
The Five Conditions That Make AI Self-Regulation Actually Work
Demis Hassabis proposed a self-regulatory organization for AI modeled on FINRA: industry-funded, federally overseen, run by independent technical experts. Frontier labs would submit models 30 days before release, benchmarks would update quarterly, and the body could coordinate slowdowns if catastrophic risks emerged. Every major lab signed on. The reception looked like consensus. David Sacks outlined five conditions he told Hassabis directly, none of which are in the current proposal. Without them, he argued, the SRO becomes the opening bid in a ratchet toward full government control. The five: first, broad representation including startups and open-source, not just the three largest labs, to prevent regulatory capture by incumbents. Second, review only true frontier models, defined as a step change above current state-of-the-art, so the leaders cannot use the process to tie up smaller competitors. Third, limit scope to catastrophic risk only, meaning cyber and CBRN (chemical, biological, radiological, nuclear), explicitly excluding disinformation and speech. Fourth, keep it voluntary until it proves it works before it becomes legally mandatory. Fifth, treat it as a substitute for a new government agency, not an addition to one. 'If it's just additive, then it defeats the purpose,' Sacks said. Without preemption written into law, he argued, this will not hold. The alternative Sacks called the FAA for AI, which Dario Amodei has repeatedly advocated. The FAA requires type certification for any new aircraft design, a process that takes 5 to 9 years for a new design and 3 to 5 years for major amendments. Applied to AI, that replaces a system releasing new model versions every few months with one requiring years of government approval per release. 'We will just simply lose the AI race if that happens because China is not going to abide by those rules,' Sacks said.
THE PLAY
If you are evaluating any AI governance proposal, run it against these five gates before supporting it: Does it include startups and open-source, or only incumbents? Does it cover only true frontier models? Is scope limited to catastrophic risk, not speech or content? Is it voluntary before mandatory? Does it explicitly preempt and replace new agencies, or stack on top of them? A proposal that fails any one of these five is not a self-regulatory solution. It is the first step toward the DMV for AI.
Token Spend Is the Next Earnings Miss Nobody Sees Coming
The PayPal Bid Is a Direct Attack on Visa and Mastercard Rails
Monitor API Token Spend by Model
The Self-Regulatory Organization Model for AI
Use Bloom Energy for Permitting Advantage
Respond to PayPal Acquisition Opportunity
Target Cosmetic Skin First for Age-Reversal Enzyme
Do Not Steal Trade Secrets from Former Employer
Build On-Device AI with Apple Silicon
Correct Data Center Myths Preemptively
Study the PJM Forward Auction Model to Understand the Energy Crunch
Subscribers Only
Unlock the Full Summary
11 more tactics + the full Action Plan
Subscribe for the Full SummaryFrom $19.99/mo · cancel anytime
Already subscribed? Log in
ALL-IN · EXTRACTED BY PODEX
