Coinbase has revealed that between 95% and 100% of its code is now written by or with the assistance of AI and large language models, according to an interview with Rob Witoff, the company’s Head of Platform. This marks a dramatic increase from just 40% five months ago in February, making it one of the most significant public disclosures of AI adoption by a major publicly listed financial technology firm.
Witoff described AI coding as effectively universal across Coinbase, with employees using LLM‑powered tools daily for drafting, refactoring, testing, reviewing, debugging, and generating boilerplate. Engineers are now running between five and ten AI agents at any given time, the output of which is estimated to be equivalent to the work of roughly 1,200 full‑time employees. However, human oversight remains central for sensitive areas like cryptography and core security systems, given the direct financial and regulatory consequences of software errors in Coinbase’s trading systems, custody infrastructure, wallets, compliance tools, and blockchain integrations.
The disclosure follows a May 2026 restructuring that cut approximately 14% of the company’s workforce (around 700 jobs). CEO Brian Armstrong explicitly linked the layoffs to AI, stating that AI had “dramatically” changed how work gets done and that Coinbase needed to return to startup speed with AI at its core. Armstrong claimed engineers were using AI to accomplish in days what previously required entire teams working for weeks.
Supporters argue that AI‑assisted development could improve Coinbase’s operating leverage, potentially boosting margins and shortening development cycles. Critics caution that the long‑term maintenance and security costs of financial infrastructure built increasingly on AI‑generated code remain untested at scale. The “vibe coding” trend—where developers accept AI output with minimal scrutiny—is not Coinbase’s model, Witoff stressed, emphasizing supervised assistance with engineers responsible for oversight and deployment.
The speed of the shift, from 40% to near‑universal AI assistance in a matter of months, highlights a broader competitive implication. A crypto exchange that can prototype, iterate, and ship with fewer employees gains a structural speed advantage in a market where product velocity correlates with user acquisition and trading volume. Coinbase’s move also puts pressure on other exchanges and infrastructure providers to adopt similar AI integration or risk falling behind in innovation.
