Armstrong Gives Coinbase Engineers a Week to Go AI-First in Coding

Coinbase CEO Brian Armstrong has issued a bold directive to engineers: adopt AI coding assistants like GitHub Copilot and Cursor within a week or face consequences, targeting 50% AI-generated code by October 2025. Announced on the “Cheeky Pint” podcast with Stripe’s John Collison, this move underscores Coinbase’s aggressive push to integrate artificial intelligence into its development core.

The AI Coding Mandate

Armstrong rejected slow adoption estimates, insisting engineers onboard AI tools immediately. In a company Slack post, he mandated, “AI is important. Onboard by week’s end, or explain why in a Saturday meeting.” Some engineers, lacking valid reasons like travel, were fired, highlighting his commitment to an AI-first culture. Currently, 40% of Coinbase’s daily code is AI-generated, up from 18% in April 2025, with a goal of surpassing 50% soon.

Why AI Matters for Coinbase

AI-driven coding accelerates development, reduces errors, and optimizes workflows, critical for Coinbase’s secure crypto platform. The company hosts monthly “AI speedruns” to train engineers, focusing on front-end development and unit testing while maintaining human oversight for security-critical systems. Armstrong noted, “You don’t want vibe-coding for systems moving money,” emphasizing rigorous code reviews.

Industry-Wide Impact

Coinbase’s strategy sets a benchmark for tech and crypto firms, aligning with leaders like Google and Microsoft pushing AI adoption. However, concerns linger about over-reliance on AI-generated code, with Armstrong acknowledging challenges in managing AI-driven codebases. Despite internal criticism for his “heavy-handed” approach, Armstrong’s vision positions Coinbase as a pioneer in AI-driven development.

Driving Innovation

By mandating AI adoption, Coinbase aims to enhance platform security, user experience, and feature deployment, reshaping its engineering culture for a competitive edge in the crypto market.