Onboard or get off board: Coinbase CEO fires engineers who refused to use AI coding tools

Coinbase CEO Brian Armstrong has mandated that all software engineers use AI coding tools like Cursor and Copilot, firing employees who failed to comply with the directive by a set deadline. The cryptocurrency exchange now generates approximately 33% of its code using AI and aims to reach 50% by the end of the quarter, while expanding AI adoption beyond engineering into decision-making processes across the company.

What you should know: Armstrong took a direct approach to enforcing AI adoption, personally meeting with engineers who hadn’t onboarded the required tools.

• Engineers who failed to use AI coding tools by the deadline were terminated, despite some resistance to what Armstrong acknowledged was a “heavy-handed approach.”
• The CEO posted in the company’s all-hands Slack channel requiring all engineers to onboard AI tools by the end of the week or attend a Saturday meeting to explain why.
• “Some people really didn’t like it, by the way, that heavy-handed approach, but I think it did set some clarity at least that we need to lean into this and learn about it,” Armstrong said.

The big picture: Coinbase is positioning AI as central to its operations, moving beyond simple coding assistance to participate in strategic business decisions.

• AI now has a designated role in the company’s RAPIDS decision-making process, contributing input alongside human participants.
• Armstrong revealed they’re “testing the limits” of when AI can “actually start to be the decision-maker on some things and do better than humans.”
• The company hosts monthly “AI Speed Run” training sessions where engineers share best practices for AI implementation.

Key details: AI adoption extends across multiple departments beyond engineering, with specific applications tailored to each team’s needs.

• Design teams are using AI heavily for creative processes, while product managers leverage it for planning and strategy.
• The finance and planning team is exploring AI for revenue forecasting by ingesting data and generating predictions.
• Armstrong emphasized that security-critical financial system code isn’t appropriate for AI generation, but front-end development can benefit from faster iteration.

Important stats: Coinbase has achieved significant AI integration in its development processes with ambitious growth targets.

• Currently, 33% of the company’s overall code is written by AI.
• The goal is to increase AI-generated code to 50% by the end of the current quarter.
• Armstrong personally uses AI extensively in his CEO role, including for decision-making processes.

What they’re saying: Armstrong provided candid details about the implementation process during his appearance on the “Cheeky Pint” podcast.

• “You’re telling me— why can’t every engineer just onboard by the end of the week?” Armstrong questioned when told adoption would take quarters to achieve.
• “I jumped on this call on Saturday and there were a couple people that had not done it… some of them didn’t [have a good reason], and they got fired.”
• “You don’t want people vibe coding these systems moving money,” he cautioned about the limits of AI-generated code in financial systems.