From 40% to Nearly 100% in Months Coinbase has disclosed that artificial intelligence now helps write between 95% and 100% of its code, a dramatic jump from the 40% figure the crypto exchange
Coinbase has disclosed that artificial intelligence now helps write between 95% and 100% of its code, a dramatic jump from the 40% figure the crypto exchange reported as recently as February. The update, shared by Head of Platform Rob Witoff, underscores how quickly the company has reorganised its engineering operations around large language models.
Witoff said that effectively every employee at Coinbase uses AI on a daily basis, with most engineers running between five and ten AI agents simultaneously. Collectively, those agents are performing the coding work of roughly 1,200 employees. Prototyping is now nearly fully automated, allowing small teams of two or three people to do work that previously required more than ten, though core cryptography still relies heavily on human review.
The speed of the shift is striking. AI-generated code accounted for just 5.7% of Coinbase's codebase in the first quarter of 2025, crossed 50% by the fourth quarter of 2025, and reached roughly 80% by early 2026 before surging to near-total coverage today.
The disclosure follows Coinbase's decision in May 2026 to cut approximately 700 employees, representing around 14% of its global workforce. CEO Brian Armstrong cited both crypto market volatility and AI's accelerating impact on how the company operates as reasons for the reduction. Armstrong framed the restructuring as a push to return to startup-level speed, with AI at the centre of how Coinbase builds and runs its products.
The company is also replacing traditional management layers with what it calls "AI-native pods," small teams, sometimes even one person, directing multiple agents that span the responsibilities of engineers, designers, and product managers. Armstrong has said the goal is to cap the leadership structure at no more than five layers below the CEO, aiming to cut what he described as the "coordination tax" created by excess hierarchy.
Looking further ahead, Witoff said Coinbase believes its AI deployment could eventually be equivalent to the output of 100,000 employees by 2030, a figure that signals the company sees its current transition as the beginning of a much larger structural shift rather than a one-time cost-cutting exercise. Coinbase has also noted that its AI spending has remained flat despite growing usage, pointing to efficiency gains in how models are being run.
Coinbase is not alone in this direction. Block, Pinterest, and Shopify have each tied recent workforce reductions or restructurings to AI adoption, reflecting a broader reckoning across the tech sector over how automation is reshaping engineering teams and headcount planning.
Sources:CoinTelegraph: Over 95% of Coinbase's Code is Now Written with AICNBC: Coinbase Cuts Headcount by 14% Citing AI AccelerationCrypto Briefing: Coinbase Says 95% to 100% of Its Code is Now AI-Assisted
