ATRIUMsearch → argument graph
ExampleVideo · 14:36 — 15:45

Stripe Minions, which execute one-shot agent-driven PRs without iteration, now produce roughly 30% of Stripe's weekly pull requests — a trajectory that points toward the future of software development.

Will details Stripe Minions — an internal tool where agents receive a single prompt, build the feature, run CI/CD, and submit a PR for human review without iteration — growing from 1,200 PRs/week to 7,000 in just months, now constituting ~30% of all Stripe PRs. ✦ AI generated

Will Gaybrick · a16z Podcast · 2026-08-17 · original ↗

starts at this moment · 14:36

Elicited by

How do you guys ship so much product?

So we created something called Stripe Minions, which we've blogged about a little bit. And, you know, of course, we've got tons of developer tooling to help users iterate with agents, but minions, we think of as, you know, the most — like one of our most important metrics internally is how many PRs and what percentage of our PRs are created by minions. And the reason for that is that minions are one shot. So you give it a prompt and you know it's going to build it and then it's going to go through CI/CD and all of testing and then you're going to review it. So you're not going to iterate, not go into planning mode. You're just going to say this is what I want, go do it. And so we think, you know, that's kind of where the world is going. They were doing 1,200 PRs per week. And last week, I think 7,000 PRs came from minions. And about 30% of our PRs in that week came from Minions.

verbatim transcript · starts at 14:36

Transcript · around this moment

14:36we've blogged about a little bit. And, you know, of course, we've got tons of developer tooling to help users iterate with agents, but minions, uh, we think of as, you know, the most like one of our most important metrics internally is how many PRs and what percentage of our PRs are created by minions. And the reason for that is that minions are one shot. >> Yeah. So you you give it a prompt and

15:02you know it's going to build it and then it's going to you know go through CI/CD and all of testing and then you're going to review it. So you're not going to iterate not go into planning mode. You're just going to say this is what I want go do it. >> And so we think you know that's kind of where the world is going. You know it's

15:18the the oneshot or the glorified real loop where you're just letting the agent be super super powerful. And so we blogged about Minions, I think in January, February, and they were doing 1,200, you know, PRs per week. And last week, I think 7,000 PRs came from minions. Wow. >> Yeah. And and about 30% of our PRs in that week came from Minions. So, >> you know, that sort of developer tooling

15:45and then the internal processes just get out of the way and make the tools amazing. And that's how we're going to go fast. >> That's amazing. In terms of empowering senior engineers who are now more powerful than ever, have you made any dramatic changes to how you organize the company internally? So, obviously that the minions example is incredible and I assume the 30% is going to go to a very

16:08high percentage within a year or something. Um, but some of the things that we've heard from other companies is taking your most powerful engineers, putting them into business units or closer to the customers. Um but then you have sort of coordination problems that come from that. So like how have you guys approached that empowerment? >> Yeah. So engineers have always been you know some of the most important product

Around this claim
This moment responds to
gives exampleStripe has evolved from a payments company with add-ons into a multi-product financial infrastructure platform focused on reducing friction and increasing agency for businesses.Will Gaybrick · a16z Podcastgives exampleAI-driven productivity gains should be used to build more and grow, not to cut headcount — the best way to optimize cost structure is to grow more.Will Gaybrick · a16z Podcastprovides contextEngineers need to learn how to reliably ship production-grade software with coding agents, including delegating work, writing executable specs and plans, and verifying AI-generated code.ByteByteGo · ByteByteGo NewslettersupportsThe entire value of AI right now is being realized by writing software. Agents are just quickly spun-up applications, and as enterprises discover they have spun up too many inefficient agents, they will realize strong software engineering capability and a strong IDE are required to fix the mess, making Cursor IDE position extremely valuable.David Friedberg · All-In PodcastsupportsEngineers need to learn how to reliably ship production-grade software with coding agents, including delegating work, writing executable specs and plans, and verifying AI-generated code.ByteByteGo · ByteByteGo Newsletterprovides contextAgentic commerce hasn't yet had its Cambrian explosion because key primitives are still missing — including machine-readable payment protocols and ephemeral micro-consumption APIs that would let agents transact without requiring humans to create accounts.Will Gaybrick · a16z PodcastsupportsAI-driven productivity gains should be used to build more and grow, not to cut headcount — the best way to optimize cost structure is to grow more.Will Gaybrick · a16z Podcast