🟣 The Collison Brothers LIVE on TBPN
TL;DR
Stripe says AI is creating more companies — and better ones — Patrick and John Collison said March was an all-time record for Stripe Atlas incorporations and Q1 new businesses joining Stripe were up 71% year-over-year, with average revenue per business rising too, not just the count of tiny “vibe-coded” startups.
The big shift is from seat-based SaaS to usage-based billing — Stripe is betting that AI products priced by tokens and inference will replace classic per-seat software, which is why it bought Metronome and is pushing real-time streaming payments, including stablecoin-based pay-per-token flows.
AI fraud is exploding alongside AI growth — John Collison said one in six new accounts across the AI ecosystem are fraudulent, with attackers spinning up free trials, reselling tokens, and even running distillation attacks, so Stripe is extending Radar beyond payments into signup and trial scoring.
Stripe wants to be the operating system for agent commerce — the company announced work with Google, OpenAI, Microsoft, and Meta so users can buy inside AI interfaces, and launched products like Stripe Projects so agents can sign up for services, buy domains, and pay for tools like Vercel or Cloudflare on a user’s behalf.
The Collisons think developer-first strategy accidentally became agent-first strategy — Stripe’s obsession with APIs, CLI tools, and documentation now maps perfectly onto how coding agents work, with John saying “agents want good DX even more than humans do.”
Stablecoins look flat only if you stare at price charts — across the later conversation with Privy, Stripe, and Bridge leaders, the recurring point was that stablecoin utility is still accelerating in cross-border payments, streaming micropayments, and agentic transactions, even if total assets have recently plateaued around the headlines.
The Breakdown
The Collisons open with one number: 10,000 people and an economy replatforming around AI
Live from Stripe Sessions, Patrick Collison frames the conference around a simple thesis: “the economy is replatforming around AI,” with 10,000 attendees there to watch it happen. The hosts needle Stripe about “controlling 1.6%” of the global economy, and Patrick’s answer is basically: the internet economy is now almost the whole economy.
Stripe’s startup data is going vertical
The most concrete signal in the interview is Stripe’s internal data: March was an all-time record for new Atlas incorporations, and Q1 new businesses joining Stripe were up 71% year-over-year. Patrick says the surprising part is that it’s not just more hobby projects — the average and median businesses are doing better too, with more small firms crossing $1 million in annual revenue.
Why the “idea guy” might be back
The brothers argue AI is making every step of entrepreneurship a little easier: incorporation, contracts, hiring, ads, engineering, design, and PM work all collapse into one high-agency person. They riff on Sam Altman’s “return of the idea guy,” but the point is serious: Stripe sees that shift directly in its onboarding numbers.
Stripe’s next act: agent commerce, usage billing, and anti-fraud
John says “all the commerce is going agentic,” and Stripe wants to sit in the middle of it. On the consumer side, it announced Google joining an agent commerce suite alongside OpenAI, Microsoft, and Meta so users can go from chatting about a bike case to just hitting buy inside the AI interface.
Patrick adds the back-end piece: AI is pushing business models from seat-based SaaS to usage-based billing, which is why Stripe bought Metronome and is building streaming payments for token-by-token billing. At the same time, fraud is getting nasty — John says one in six new accounts across the AI ecosystem are fraudulent — so Stripe is extending Radar to score signups, free trials, and token abuse, comparing it to fighting “shrinkage” so the internet doesn’t turn into a CVS with everything locked up.
From payments company to “home” for money
Before leaving, Patrick lays out the broader ambition: Stripe should not just move money through a pipe, but become “a home.” That means Treasury, balances held inside Stripe, global payouts to 150-plus countries, and a programmable financial-services layer that turns Stripe into an operating system for money movement.
John Collison on Stuart Brand, OpenClaw, and why tinkering matters again
After Patrick exits, the conversation gets looser and more cultural. John connects Stuart Brand, the Whole Earth Catalog, and the “mother of all demos” to today’s AI moment, saying he’s energized by tools that create individual empowerment, not passive “slop.” He defends the OpenClaw/Mac mini tinkering scene as a return of Homebrew Computer Club energy, and seems genuinely baffled that Silicon Valley claims to miss hacker culture and then mocks it when it comes back.
AI software is messy now, but John thinks the doomers are “high honestly”
Asked about tool-shaped objects, maintenance debt, and the “death of software engineering,” John says we’re clearly in a transitional phase where prompting is still awkward and brittle. But he rejects the idea that software engineers are obsolete, arguing that people who understand what’s happening under the hood — model, memory, harnesses, abstractions — will be more valuable, not less, and that high-agency people plus “double majors” in software and another domain will crush the next decade.
The back half turns into a stablecoin and Stripe infrastructure jam session
With Privy, Jeff Weinstein, and Bridge’s Zach Abrams, the broader Stripe orbit makes the same case from different angles: stablecoins aren’t dead, they’re becoming useful. Jeff says Atlas now incorporates 26% of all new Delaware C-corps and hints at an “Atlas CLI,” while Stripe’s developer-first tools are increasingly being consumed by agents, not humans.
Zach from Bridge says stablecoin supply may look flat in headlines, but transaction activity, builder interest, and real usage are still climbing fast — especially in cross-border payments and future agentic systems. The shared vibe across all three conversations is that boring infrastructure is suddenly becoming the most important thing in AI: billing, permissions, compliance, payout rails, and all the unglamorous stuff that lets autonomous software actually do work in the world.