Coinbase Chief Executive Officer Brian Armstrong says some engineers were fired for not adopting AI tools.
He spoke about the policy on John Collison’s podcast, Cheeky Pint, last week. The message was direct and clear: Use AI. Move fast. Or step aside.
Armstrong said adoption had moved too slowly. Teams planned a gradual rollout, but he disagreed.
He pushed for immediate onboarding to Cursor and GitHub Copilot. He posted in the company’s all-hands Slack.
He set a one-week deadline. He scheduled a Saturday call with anyone who still had not onboarded AI.
Some engineers had valid reasons; they were traveling or had just returned from a trip. Others did not.
Those without justification were dismissed. Armstrong called the approach “heavy-handed.” He also said it brought clarity. The company had to lean in and learn.
Coinbase
Armstrong said about 33% of Coinbase’s code now comes from AI. The target is 50% by the end of the quarter.
Not all code is equal, though. Systems that move money demand strict review. Therefore, he warned against “vibe coding” core financial rails.
Yet front-end work is different. There, AI speeds iteration, which helps teams ship faster.
Training And Guardrails
Coinbase runs monthly “AI Speed Runs.” An engineer leads each session; they show real workflows, they share prompts, and they compare outputs.
The goal is to spread good practice; human code review remains in the loop. Security checks stay in place.
AI In Management
Coinbase uses a process called RAPIDS for major decisions, and each stakeholder adds input.
Armstrong added a new row: AI. The system now submits its own written view for review next to human input.
Transcript Excerpt
John Collison: What are other ways in which Coinbase is crypto-pilled, AI-pilled, works differently from a company founded 10 or 20 years prior?
Brian Armstrong: Like a lot of companies, we’re leaning as hard as we can into AI.
John Collison: What does that mean, concretely?
Brian Armstrong: We’re doing a lot of the best practices. We made a big push to get every engineer on Cursor and Copilot. Then another question was, “Well, are they actually going to use it?”
John Collison: Was it you or Tobi [Lütke, member of Coinbase board] who mandated it?
Brian Armstrong: I mandated it. Yeah.
John Collison: You required people to have a call with you who-
Brian Armstrong: That’s true. I did do that.
John Collison: You required people to justify to you, the CEO, if they weren’t using AI code.
Brian Armstrong: That’s true. Originally-
John Collison: Sorry, maybe I’m not meant to tell that story?
Brian Armstrong: No, I don’t mind. It’s actually a good story. Originally they were coming back and saying, “Over the next quarter, two quarters, we’re going to get to 50% adoption.” I said, “Why can’t every engineer just onboard by the end of the week?” So I posted in the all-in Slack channel.
John Collison: Just a light dusting of founder mode.
Brian Armstrong: I said, “AI’s important. We need you all to learn it and at least onboard it… by the end of the week. And if not, I’m hosting a meeting on Saturday with everybody who hasn’t done it.”
I jumped on this call on Saturday and there were a couple people that had not done it. Some had good reasons. Some didn’t, and they got fired. Some people really didn’t like that heavy-handed approach, but it set clarity.
John Collison: What’s your experience of AI coding been so far?
Brian Armstrong: We’re still figuring that out. One thing we started doing is a monthly AI Speed Run. We’re doing about 33% of code written by AI now. We have a goal to get to 50% by the end of the quarter. You probably can go too far with it.
You don’t want people vibe coding these systems moving money. You have to code review it and keep humans in the loop. But for some front-end pieces, you can iterate faster.
Brian Armstrong: It should not be just engineering. Design is using it heavily. Product managers. I think FP&A could use it to ingest data and forecast revenue. We use a decision process called RAPIDS.
We now have a row for AI that writes its input as one of the people that help make decisions. We’re testing the limits—when it can be the decision-maker and do better than humans.
