BREAKING: John Collison… | Sourcery Newsletter Summary

1. Key Themes

Theme: Agents will transact on the existing web, so computer use matters more than new agent-specific infrastructure

Computer use as the "backwards compatibility layer"

John Collison argues that mature computer use lets agents act on the internet as it exists today, without waiting for every site to build an API or MCP server.

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

"Instead of waiting for every organization to build an API or MCP server, agents can interact with the websites and software that already exist."

Web forms are the first thing agents eliminate

The example is renewing a driver's license: an agent holding your address and payment details navigates the DMV site for you.

"Filling out a web form is not a value-add activity for anyone."

The browser as the agent's primitive

Browserbase CEO Paul Klein IV makes the same bet from the infrastructure side.

"We aren't gonna need to rebuild the web for AI." / "we might just get away with AI using a browser just like we do."

"If you want AI to do the same work that you do on the web, you need to give it a browser."

Theme: AI becomes the new point of aggregation for commerce, which may favor the long tail

AI discovery as the new aggregator

Collison says AI-powered product research changes where purchase decisions begin, and incumbent aggregators will have to adapt.

"AI discovery is the new point of aggregation for commerce."

Smaller brands get a better shot

"My experience of doing AI-powered product research is that it often surfaces… [niche brands]" (from Molly O'Shea's tweet), and in the article: "I think it's quite a good story for the long tail of businesses."

Well-reviewed, well-liked products become more discoverable regardless of marketing scale.

Theme: Stripe is rebuilding and acquiring for an agent- and AI-native economy

Scale and momentum

"Stripe processed $1.9T in 2025, up 34%. Now the company is rebuilding for an internet where AI agents, not humans, increasingly transact."

Acquisitions map to the AI stack

Stripe bought stablecoin/crypto infrastructure (Bridge, Privy), usage-based billing (Metronome), and multi-model routing (OpenRouter).

"Bridge + Privy, bringing crypto-native infrastructure & talent / Metronome, usage-based billing for the AI era / OpenRouter, routing across multiple AI models as inference spend & model competition explodes"

The reported price tag is large: "Stripe has reportedly spent ~$9.6B+ acquiring Bridge, Metronome & OpenRouter, plus an undisclosed amount for Privy."

AI is visible in new company formation

"Stripe is seeing its highest rate of new firm creation in recent memory, while its 2026 Atlas cohort is tracking at 5x the revenue of the 2025 class."

Theme: Multi-model usage is the thesis behind model routing

Why OpenRouter

"To believe in OpenRouter, you have to believe that more than one model will be used inside a business."

Businesses may use different models for different workloads and swap them as new models arrive. Collison's example: route a high-value user to a frontier model and a lower-value user to a cheaper one. Stripe judged OpenRouter furthest ahead in building the marketplace and infrastructure for this.

2. Contrarian Perspectives

Don't rebuild the web for agents, just give them a browser

The consensus view is that every service needs APIs and MCP servers. Collison and Klein argue the opposite: agents will use the human web.

"Some people say, 'Well, we need to add MCP everywhere, or APIs,' but actually, we might just get away with AI using a browser just like we do." (Paul Klein IV)

Supporting reasoning: waiting on every organization to build agent interfaces is a bottleneck that computer use bypasses.

Developer ergonomics are not AI ergonomics

Stripe spent years polishing APIs, docs, and integrations for human developers, but that investment doesn't transfer one-to-one.

"No one actually really integrates the Stripe API by hand anymore. You know, they send their coding agent to do it."

"The developer ergonomics are different from the AI ergonomics."

Evidence: Stripe now has to address problems like agents integrating the wrong version of an API.

AI as a boost to the long tail, not just incumbents

Many expect AI to concentrate power in big platforms. Collison sees AI discovery surfacing lesser-known brands and weakening existing aggregators' hold.

"I think it's quite a good story for the long tail of businesses."

3. Companies Identified

Stripe

Description: Global payments infrastructure company, co-founded by John Collison.
Why mentioned: Central subject; rebuilding its APIs and business for agentic commerce.
Quotes: "Stripe processed $1.9T in 2025, up 34%." / "Stripe considers 2026 the beginning of the 'singularity epoch.'"

Browserbase

Description: Cloud browser infrastructure for AI agents; framework Stagehand; hosted the Navigate event.
Why mentioned: Infrastructure bet behind agents using the existing web.
Quotes: "Browserbase is building the infrastructure for this shift, giving AI agents cloud browsers to navigate the existing web & take actions across it." Raised a $40M Series B (June 2025 per linked post).

OpenRouter

Description: Multi-model routing and marketplace for AI models.
Why mentioned: Stripe's largest acquisition in the group (~$7.5B+ reported; ~$8B in one tweet).
Quotes: "Stripe saw OpenRouter as furthest ahead in building the marketplace and infrastructure for businesses using multiple models."

Bridge

Description: Stablecoin infrastructure.
Why mentioned: Stripe acquisition (~$1.1B reported).
Quotes: "Bridge + Privy, bringing crypto-native infrastructure & talent."

Privy

Description: Crypto wallet infrastructure.
Why mentioned: Stripe acquisition (undisclosed).
Quotes: "Privy (Undisclosed) for crypto wallets."

Metronome

Description: Usage-based billing platform.
Why mentioned: Stripe acquisition (~$1B reported) for AI-era pricing.
Quotes: "Metronome, usage-based billing for the AI era."

Hyperagent

Description: Agent company founded by Howie Liu (previously Airtable).
Why mentioned: Founder predicts a massive agent market.
Quotes: AI agents will become "a multi-deca-trillion-dollar market."

Keycard

Description: AI agent identity/authorization company led by Ian Livingstone.
Why mentioned: CEO argues autonomy is blocked by something other than intelligence.
Quotes: "We have apparently AGI, but the real thing that stops us from having autonomous agents that actually work..." (truncated in source).

Brex

Description: Finance platform (cards, expenses, travel, bill pay, banking).
Why mentioned: Newsletter sponsor.
Quotes: "Trusted by OpenAI, Anthropic, Vercel, Granola, Deepgram, & Sourcery."

Airtable / Bending Spoons

Description: Howie Liu's prior company and its acquirer.
Why mentioned: Background on Liu.
Quotes: "prev Founder & CEO of Airtable, acq. by Bending Spoons."

4. People Identified

John Collison

Description: Co-founder, Stripe.
Why mentioned: Interview subject on agentic commerce, computer use, and Stripe's AI rebuild.
Quotes: "Computer use is the ultimate backwards compatibility layer for the real world." / "Who said I don't have AI psychosis?"

Paul Klein IV

Description: Founder & CEO, Browserbase.
Why mentioned: Host of Navigate; proponent of agents using browsers.
Quotes: "We provide that browser, at scale, as infrastructure."

Howie Liu

Description: Founder & CEO, Hyperagent; former Airtable CEO.
Why mentioned: Hot take on agent market size.
Quotes: "The agent opportunity is going to be trillions of dollars."

Ian Livingstone

Description: CEO, Keycard.
Why mentioned: Argues intelligence isn't the bottleneck for autonomous agents.
Quotes: "We have apparently AGI, but the real thing that stops us from having autonomous agents that actually work..."

Erika Bricky

Description: Browserbase team member.
Why mentioned: Credited for organizing the event.
Quotes: "Big shoutout to Founder & CEO Paul Klein IV & Erika Bricky for such a top notch event."

Molly O'Shea

Description: Author and interviewer, Sourcery.
Why mentioned: Conducted the conversation.
Quotes: "I sat down with John at Browserbase Navigate."

5. Operating Insights

Optimize your product for AI ergonomics, not just developer ergonomics

If coding agents are now your primary integrators, documentation and APIs need to be designed for them, including versioning pitfalls.

"Stripe is now having to change what it optimizes for internally, including problems like agents integrating the wrong version of an API."

Design pricing and infrastructure for multi-model, usage-based AI workloads

Route different users or workloads to different models by value and cost, and expect to swap models as better ones appear.

"sending a high-value user to a frontier model while routing a lower-value user to a lower-cost model."

Stay current by building and getting outside the bubble

Collison keeps up by vibe coding tools he wants and spending time with fast-growing customers, then deliberately disconnecting.

"Putting down the phone from time to time. I don't know if people have tried that, but it's pretty nice out there in the wider world."

6. Overlooked Insights

Good bots vs. bad bots is a coming problem

The timestamps list "The Coming Good Bot vs. Bad Bot Problem" and "Why AI Could Mean More Cyberattacks," implying agent traffic will force new trust, identity, and fraud infrastructure, an angle Keycard's focus on agent autonomy also touches.

"([23:28]) The Coming Good Bot vs. Bad Bot Problem" / "([04:40]) Why AI Could Mean More Cyberattacks"

Stripe frames 2026 as a "singularity epoch"

A striking internal framing for how Stripe views the current moment, which likely shapes its aggressive acquisition pace.

"why Stripe considers 2026 the beginning of the 'singularity epoch.'"