Key facts
- Over 95% of Coinbase's code is now written with or by large language models (LLMs).
- This represents a significant increase from the 40% reported in February.
- Coinbase CEO Brian Armstrong emphasized AI's role in accelerating work pace.
- The company previously laid off 14% of its workforce in May, partly due to AI efficiencies.
- AI agents are estimated to perform the work equivalent of 1,200 employees currently.
Coinbase is now leveraging artificial intelligence to write between 95% and 100% of its code, a substantial increase from the 40% reported in February. Rob Witoff, Coinbase's head of platform, stated that this accelerated adoption of LLMs is transforming the company's operations, enabling smaller, more senior teams to achieve greater productivity.
Witoff explained that while core cryptography still requires meticulous human oversight from industry-leading cryptographers, tasks like building internal prototypes are now fully automated. AI is also extensively used for testing code to ensure functionality and identify vulnerabilities. This shift has allowed Coinbase to reorganize into smaller teams, where a few employees can now accomplish work previously requiring ten or more.
This increased reliance on AI follows Coinbase's decision to cut 14% of its workforce, approximately 700 employees, in May. CEO Brian Armstrong cited AI's dramatic impact on the pace of work as a reason for needing to return to the efficiency of the company's startup days. Witoff noted that most Coinbase engineers currently manage five to ten AI agents, collectively performing the coding equivalent of about 1,200 employees. He projected that by 2030, AI agents could be doing the work of 100,000 employees.
Coinbase is not alone in this trend; other crypto companies such as Crypto.com, Block, Kraken, Gemini, Messari, and Dune have also cited AI-related efficiencies in their recent workforce reductions.