John Collison, Stripe: AI Agents Will Rewire the Internet

1. Key Themes

Computer use is the "backwards compatibility layer" for the real world

Collison argues the next hop in AI progress is mature computer use, because it lets agents operate any website without waiting for institutions to build agent-native interfaces. This is the thesis behind his enthusiasm for BrowserBase, the host of the event. He describes it as a step-change comparable to transformers, LLMs, and RL.

• "computer use is the ultimate backwards compatibility layer for the real world. Like how long would you need to wait for, you know, the DMV to add an MCP? Like I'm not a betting man, but I'm betting we would have to wait a few years. But it doesn't matter because computer use solves that." — John Collison 00:01:55
• "my perspective, at least, is that good computer use is, like, will come to be viewed as one of these next hops. Because it is the backwards compatibility layer on the rest of the world." — John Collison 00:20:03

Agentic commerce makes AI discovery the new point of aggregation

Collison frames agentic commerce as a two-order shift. First, agents do a huge share of purchasing. Second, discovery moves to AI, which reshuffles which merchants win. The long tail benefits because AI research surfaces well-reviewed niche brands, and incumbent aggregators must answer for that.

• "my experience of doing AI powered product research is it often actually surfaces niche brands or businesses you haven't heard of. Like I think it's quite a good story for the long tail of businesses." — John Collison 00:02:52
• "AI discovery is the new point of aggregation for commerce. And so you probably have a bit of a turning over of the top businesses that's going to happen" — John Collison 00:03:18
• "Second order, it will probably actually reshape the product and the merchant landscape to favor the people who can most successfully adopt it." — John Collison 00:04:30

The Facebook/Instagram-ads analogy: native adopters beat incumbents

The historical parallel Collison offers is that the last generation of ad targeting created new companies native to the platform that out-acquired big-box retailers who had the same tools. He expects AI commerce to do the same, with consequences flowing all the way down to product creation.

• "Think of a wish.com or a teespring or, you know, even Timu and Sheehan, you know, to a significant degree, grew up on those platforms and did a huge amount of customer acquisition on those platforms." — John Collison 00:03:39
• "they were able to more natively adopt us than the prior generation of big box retailers who in theory had all the same tools available to them, but they were just less native" — John Collison 00:04:08

The AI-native economy: new firm creation is surging

Stripe's payments data offers a leading indicator that cheaper intelligence is increasing company formation.

• "as a result of intelligence on tap getting much cheaper, new firm creation is way up. We're seeing the highest rates of new firm creation that we've seen in recent memory on Stripe, like, right now." — John Collison 00:08:24
• Stripe processed $1.9 trillion in 2025, up 34%, per the host's opening framing. 00:00:01

Cybersecurity: offense and defense both accelerate, and legacy stacks lose

Collison expects more breaches over the next five years than the last five, because capable cyber models are available to attackers and many organizations will adopt too slowly. Startups with small, homogeneous stacks are advantaged.

• "it does feel like, say, over the next five years, you're going to see more breaches than you did over the past five years." — John Collison 00:05:38
• "startups are disproportionately in the good spot because it is easier for them to adopt the latest models with kind of the latest cyber capabilities." — John Collison 00:06:08
• "if you're running some, like, 50-year-old tech stack that has grown through acquisition... you're running on versions of Windows that aren't even supported anymore. Like, that's not a great place to be in from a cyber point of view." — John Collison 00:06:30

Everyone's AI-era chessboard got jiggled, including Stripe's own moat

Stripe's famous developer-experience obsession has been partly invalidated, because coding agents now do the integrating and have different "ergonomics" than humans.

• "Overnight, you know, relatively overnight, no one actually really integrates the Stripe API by hand anymore. You know, they send their coding agent to do it." — John Collison 00:10:07
• "the developer ergonomics are different from the AI ergonomics. And so many of the things that we spend time on to make things nice for developers actually don't really matter to the AIs." — John Collison 00:10:33
• "we are having to kind of totally change what we're good at internally at Stripe." — John Collison 00:10:33

Stripe's acquisition map follows the new stack: crypto, usage billing, model routing

Each acquisition fills a "new muscle" in a fast-moving market.

• On Bridge and Privy: "the folks on those teams are crypto native in a way that it would be hard to get there that quickly at Stripe." — John Collison 00:11:54
• On Metronome: "everyone needs to have usage-based billing these days with AI products." — John Collison 00:12:22
• On OpenRouter: "to believe in OpenRouter, you have to believe that more than one model will be used inside a business" — John Collison 00:13:17

Personalized software finally arrives, built bottom-up inside companies

Once the internal AI tool has proper data access and permissioning, employees build their own applications rather than waiting for central teams.

• "they are building their own application... someone on the legal team builds their agent for ingesting all the new regulation that comes along" — John Collison 00:18:41
• "people have been talking about personalized software for decades. And there's something about this moment in time where finally the barrier to entry has gotten low enough" — John Collison 00:19:35

The "singularity" framed as optimistic takeoff, with recursive self-improvement already visible

Stripe tongue-in-cheek declared January 1, 2026 as the singularity epoch, a nod to Unix time.

• "we're clearly in a world where AI capabilities are helping with the rate of improvement in AI. I mean, just look at all the math announcements that we had this week and last week. We're in math wars now." — John Collison 00:07:53
• "we don't see the singularity in any kind of a pessimistic or kind of worried manner. It's just like we are in this moment of AI takeoff right now and you should enjoy it." — John Collison 00:09:20

Compounding stick-with-it-ness is the underrated founder strategy

Stripe's growth, he says, never had a breakout moment; it was years of compounding, patient reputation-building, and users-first product decisions.

• "Stripe never had a South by Southwest breakout moment... every year has been a bigger year than the prior year. It turns out you just, like, do that enough years one after another and you eventually get to $1.9 trillion." — John Collison 00:25:01
• "there's a real element of Stripe that is just stick-with-it-ness where we didn't quit and we didn't get acquired and we just stayed plugging away at it. And I think that's underrated in a lot of tech stories." — John Collison 00:25:54

2. Contrarian Perspectives

Bots will be the majority of web traffic, so "bot detection" is the wrong frame

Most of the industry treats bots as a threat to filter. Collison suggests the premise is dissolving, and the real question becomes distinguishing good bots from bad ones.

• "many sites do bot detection and bot prevention, and that is a reasonable thing to do... but, like, it's increasingly all bots." — John Collison 00:24:05
• "what is the new equivalent of... bot detection for the world we are going to, where there is the, you know, the good bots and then there are the bad bots and you need to be able to screen between them?" — John Collison 00:24:05

Stripe's core differentiator, developer experience, is being commoditized by agents

It is unusual for a leader to say his company's decade-plus craft advantage is partly obsolete. API docs and migration ergonomics were Stripe's brand, yet coding agents care about different things, such as integrating the wrong API version.

• "many of the things that we spend time on to make things nice for developers actually don't really matter to the AIs. And what matters is a fairly different set of things." — John Collison 00:10:33

AI commerce will favor the long tail over today's aggregators

The default assumption is that AI platforms concentrate power in a few giants. Collison argues that agentic research rewards well-reviewed small merchants and reshuffles the rankings.

• "if you are selling a really good product, if you're making it discoverable, then you can kind of shoot up in the rankings." — John Collison 00:03:18

The AI doom narrative is misplaced; worry specifically, not existentially

Collison rejects the "standard San Francisco doom" and replaces it with narrow, specific concerns, such as whether your cybersecurity holds up, plus excitement.

• "I don't think people should feel this, like, ominous, oppressive, overbearing sense of foreboding generally. I think they should feel pretty excited to be living in this fast takeoff world." — John Collison 00:11:26

Multi-model is the future inside a single company, which makes routing a real layer

Counter to the idea that one winning model will serve everything, Collison bets on heterogeneity, with routing based on workload and user lifetime value.

• "a really high-value user you can send to a frontier model. But a lower-value user, you route to a lower-value model. And so it depends on the user's lifetime value to you" — John Collison 00:13:45

3. Companies Identified

Stripe

Payments and financial infrastructure company, 17 years old. Mentioned as the subject itself: processed $1.9 trillion in 2025, up 34%, and is retooling for agentic commerce, stablecoins, and AI billing.

• "it turns out you just, like, do that enough years one after another and you eventually get to $1.9 trillion." — John Collison 00:25:01

BrowserBase

Browser infrastructure for AI agents; the event host. Collison cites it as at the core of agents using computers and as a Stripe customer "well on its way to becoming a huge Stripe company."

• "someone like a, you know, browser base is well on its way to becoming a huge Stripe company or lots of other kind of early-stage startups like that." — John Collison 00:21:19
• "the browser bases of the world running on Stripe, clearly at the core of the AI revolution, allowing agents to actually use computers." — John Collison 00:14:21

OpenRouter

Marketplace/router for multiple AI models, acquired by Stripe. Mentioned as the furthest-ahead provider for multi-model routing.

• "as far as we could tell, they were the furthest ahead in delivering that marketplace to help businesses actually use multiple models." — John Collison 00:13:45

Metronome

Usage-based billing company, acquired by Stripe. Mentioned because AI inference costs force usage-based pricing on nearly every AI product.

• "Because of AI, everyone has inference costs these days. And so everyone has to actually price their product differently." — John Collison 00:12:22

Bridge

Stablecoin infrastructure company, acquired by Stripe. Mentioned as bringing crypto-native culture into Stripe.

• "Bridge and Privy were in crypto. And again, there is a certain kind of cryptonativity and way of working in the crypto world that is different" — John Collison 00:11:54

Privy

Crypto wallet infrastructure company, acquired by Stripe. Mentioned alongside Bridge for its crypto-native team.

• "they've injected a lot of that cryptonativity in the Stripe culture." — John Collison 00:11:54

Meta

Social media and AI company; a Stripe customer. Cited as a pre-AI company increasingly becoming an AI company, and for its Muse assistant.

• "Meta runs on Stripe. Grew up in the pre-AI era, pretty increasingly becoming an AI company." — John Collison 00:14:49

NVIDIA

GPU maker. Cited as the archetype of patient compounding and incremental improvement (such as CUDA) rather than a breakout moment.

• "NVIDIA is the largest company in the world. And they've just been making GPUs since the 1990s... it's a real story of stick-with-it-ness" — John Collison 00:26:22

Wish.com, Teespring, Temu, and Shein

Examples of companies native to Facebook/Instagram ad targeting who out-acquired big-box retailers. Cited as the analogy for the AI-commerce era.

• "Think of a wish.com or a teespring or, you know, even Timu and Sheehan, you know, to a significant degree, grew up on those platforms" — John Collison 00:03:39

Box

Enterprise content management company; led by Aaron Levy. Mentioned only for a product-prioritization conversation.

• "I remember asking Aaron Levy of Box at one point, asking how product prioritization should work." — John Collison 00:15:22

Instinct, GrokBot, and Muse (Meta)

AI personal-assistant tools Collison urges people to try for a sense of "what is coming" (names as transcribed; product names are uncertain).

• "you really should play with one of, you know, Instinct or GrokBot or Muse from Meta or any of these kind of AI assistant tools because they're a very powerful sense of what is coming" — John Collison 00:24:29

ChatGPT and Claude

Cited as examples of AI products with usage caps and credit purchases.

• "You can sign up for ChatGPT or Claude, and they will let you go a certain amount. But then they're like, sorry, you ran out of plan. You've got to buy credits." — John Collison 00:12:22

Dreamforce (Salesforce)

Mentioned only as the concurrent event crowding San Francisco.

• "especially while Dreamforce is on. I mean, good God." — John Collison 00:22:48

4. People Identified

Paul (BrowserBase)

BrowserBase leader and event host, thanked at close and opening for hosting. Mentioned for enabling the event; BrowserBase is the highlighted infrastructure company.

• "thank you to Paul and everyone at BrowserBase and all of you for coming today." — Molly O'Shea 00:27:46

Aaron Levy

Co-founder and CEO of Box. Collison recalls early-Stripe advice-seeking from him on product prioritization, with a pragmatic, if unspecific, answer.

• "he was like, yeah, you make a big list of all the product things you want to do, and then you do the things that will be most impactful." — John Collison 00:15:22

5. Operating Insights

Give your internal AI full data access with permissioning, then let employees build their own apps

The unlock at Stripe was not a better chatbot but a secure integration with internal data (with access controls and privacy guardrails), after which sales and legal teams self-built their tools. The permissioning is the hard part.

• "you give the AI application access to the internal data... you need to have all the privacy guardrails. You need to have all the access control guardrails where, like, some people need to have access to the HR database, but, like, everyone should not have access" — John Collison 00:17:49
• "all our sales teams just completely change their workflows where they can self-serve all the stuff they need to" — John Collison 00:18:17

Buy when the market is new and fast-evolving; build when it's core

Collison's build-versus-buy heuristic: build what's core to Stripe; buy when it's a brand new muscle and the market will change before your internal effort ships.

• "by the time you build your internal efforts, it'll be a little bit out of date or something." — John Collison 00:16:42
• "something that is a brand new muscle, like token routing, I trust that Stripe could... learn to get good at that. But it would take much longer." — John Collison 00:16:18

Watch your fastest-growing customers to see the future early

Collison keeps his information diet fresh not by reading papers but by spending time with fast-growing companies on Stripe, which are invariably AI-adjacent, and by vibe-coding his own tools.

• "we've always tried to, you know, look at where the puck is going rather than, you know, companies' current scale." — John Collison 00:21:19

Start from what the customer wants and avoid internally oriented thinking

Stripe's first operating principle is users first, with a warning about drifting into spreadsheet-driven market selection.

• "I think it's very easy for companies to go adrift and start getting into, you know, you know what market we should get into or, you know, a great new revenue pool would be X or, like, really get into this internal company-oriented way of thinking." — John Collison 00:27:19

Treat being a good counterparty as a strategic asset

• "we care about Stripe having a reputation for being a good company to work with. And so I think just treating being a good counterparty seriously or something is something that really matters." — John Collison 00:26:51

6. Overlooked Insights

Pricing is a new bottleneck: nearly every AI product now has COGS and overage

Said in passing while explaining Metronome, this implies the entire SaaS flat-seat pricing model is breaking, creating a large market for usage-based billing, metering, and credit systems.

• "There are very few products these days that do not have some usage-based component or some overage." — John Collison 00:12:22

AI inference is becoming a large and growing share of business spend, which sets up a new spend-management layer

The OpenRouter rationale rests on this brief observation: inference spend is now meaningful across R&D and product delivery, which makes optimization, routing, and financial infrastructure around tokens a major emerging category, especially for a payments company.

• "an increasing fraction and increasingly huge fraction of businesses' spend is actually going towards just AI and inference." — John Collison 00:12:51