Loading video...
Video Failed to Load
NEW: Stripe Co-Founder John Collison (John Collison) says AI agents will completely rewire internet commerce. Stripe processed $1.9T in 2025, up 34%. Now the company is rebuilding for an internet where AI agents, not humans, increasingly transact. “Instinct is real, Muse is real. All these things are real.” “Computer... show more
518,524 views • 6 days ago •via X (Twitter)
26 Comments
@stripe
@stripe Listen on Spotify: Apple Podcasts: Watch on Youtube:
@collision @stripe Agreed!
@collision @stripe Computer use may be the practical bridge here, but it also makes the existing web’s security assumptions much harder to preserve.
@sourceryy @collision @stripe Reimagining the future again @collision 👌
@collision @stripe Collinsons are playing 4d chess. When they inevitably ipo they will will control so much of thr web and commerce that regulators will scratch their heads over how they let it happen.
@collision @stripe The merchant side of this is less glamorous: an agent can only recommend what it can read. A lot of product data was written for humans looking at photos, not for an agent comparing sizes, materials and return windows.
@collision @stripe Really great interview. I do think that infrastructure for Agents will include an integration of payments and authentication infrastructure. When we started Storyful it was about verifying the real rather than debunking the fake, the same approach is needed now for Agents.
@collision @stripe Computer use gives agents reach; structured offers give them a basis for comparison. In fashion, two pages can show the same item but different stock, delivery costs and return terms. The useful handoff is from browsing pages to comparing fresh, equivalent offers.
@collision @stripe Who gets the final “yes, spend my money” button in this future?
@collision @stripe shoes...
@collision @stripe Stripe’s $1.9T while rebuilding for agents is the real signal. Computer use bridges the old web so agents can act.
Collison is right that agents rewire commerce. Stripe rebuilding the rail is the visible half. The half that decides who wins is what those agents negotiate once they can transact. Humans collapse onto price. Machines do not. They can clear price, speed, quality, priority and the reversal window on every job, then renegotiate all five. A rail built for agents is not the same as a market still priced for humans. If both sides are software, is the cheapest supplier even in the auction anymore? 👇
@collision @stripe What impressed me most is his ability to stay on point with that red dress. Molly is making life hard for her interview partners. 💃🏻
@collision @stripe What if the agent never has to use a UI/website to buy and pay. Instead executing the discovery and purchase through MCP/APIs. That is what I believe the future looks like, so making that possible @Kiflyai .
@collision @stripe Great post
@collision @stripe Curious how this plays in b2b. Most of the companies I build for still run purchasing through a PO and a manager signing off. I mean the payment is the easy part.. the approval chain is where agents get stuck.
@collision @stripe payments shall be..
@collision @stripe Agents transacting is the easy part. The hard part is who sets the budget. Every agent that can spend needs a prepaid limit, a named human owner, and a kill switch. Metronome and OpenRouter in the same breath says Stripe sees that bill coming.
@collision @stripe Agents as artifacts
@collision @stripe The computer-use line is the practical part. If an agent can work any screen a person can, your checkout no longer needs an API to be agent-ready, but it does need clear labels, stable buttons, and a confirm step an agent can read before money moves.
@collision @stripe If computer use really is the backwards compatibility layer, the hard part shifts to trust: agents need scoped payment permissions and clear audit trails before merchants will let them check out unsupervised.
AI agents changing how online commerce works could become one of the biggest technology shifts of the next decade. If autonomous agents start handling purchases, payments, and transactions, companies like $STRP could play an important role in building the infrastructure behind this new digital economy. The key will be adoption — turning the AI agent vision into real consumer and business usage. Curious what @NYKrant thinks.
@collision @stripe the hard part won’t be getting an agent to click buy. it’ll be handling stock changes, substitutions, refunds, and disputes clearly enough that merchants and customers trust the transaction.
@collision @stripe Backwards compatibility is a good way to put it. Most small shops will never build anything for agents, so a bot clicking their booking form is how they get the Tuesday table at all. Curious if the shop ever finds out it was an agent that booked.
@collision @stripe the part about agents replacing web forms is the only thing that actually excites me. everything else is just rich men explaining why we should trust them with more infrastructure.
