Brian Armstrong Fired Coinbase Engineers Over AI Snub

Coinbase CEO Brian Armstrong revealed this week on John Collison’s “Cheeky Pint” podcast that he promptly fired engineers who refused to test company-provided AI coding tools like GitHub Copilot and Cursor.

The Mandate: Adopt AI or Explain Why

While AI coding assistants have become industry standard for handling repetitive tasks, some developers remain hesitant. After Coinbase purchased enterprise licenses for GitHub Copilot and Cursor to cover its entire engineering team, leadership warned Armstrong that widespread adoption could take months. Skeptical of this slow timeline, the CEO decided to bypass traditional channels.

Armstrong posted a direct mandate in Coinbase’s main engineering Slack channel, emphasizing the critical importance of AI. He instructed all engineers to onboard by the end of the week, clarifying that daily usage wasn’t required yet, but initial setup was mandatory. He warned that anyone who failed to comply would have to attend a Saturday meeting to explain why.

A Saturday Meeting and Instant Dismissals

During the Saturday follow-up meeting, Armstrong confronted the remaining employees who had not set up their accounts. While some engineers offered legitimate reasons, such as returning from vacation, others lacked justification.

Those without a valid excuse were fired on the spot. Armstrong admitted this was a “heavy-handed approach” that created some internal friction, but he emphasized that it successfully communicated that AI integration is non-negotiable at Coinbase.

var playerInstance_jwplayer_6a6913b41ce4a = jwplayer( “jwplayer_6a6913b41ce4a” );
playerInstance_jwplayer_6a6913b41ce4a.setup({
playlist: “https://cdn.jwplayer.com/v2/media/nsQAyeWN”,
});

Coinbase did not respond to requests for comment regarding the exact number of terminations. To maintain momentum, the company now hosts monthly cross-team meetings where engineers share innovative ways they have integrated AI into their workflows.

The Challenges of Managing AI-Generated Code

Despite his aggressive push for adoption, Armstrong acknowledges the complexities of relying heavily on automated programming. During the podcast, Collison—who has been programming since childhood—questioned the long-term viability of managing an AI-generated codebase, noting that maintaining such infrastructure remains highly complex.

Armstrong agreed with Collison’s concerns. This caution is shared across the industry; a former OpenAI engineer previously characterized that company’s central code repository as “a bit of a dumping ground,” prompting management to redirect engineering resources toward cleanup efforts.