Points of Focus

• Coinbase said AI now helps write 95%-100% of its code, up from 40% in February.
• Engineers now work with multiple AI agents as AI becomes central to software development.
• Human oversight remains in place for cryptography and other security-critical systems.

Coinbase said AI has become a core part of its software development process, with AI now helping write between 95% and 100% of the company’s code. The disclosure shows how quickly Coinbase has expanded AI-assisted software development across its engineering teams.

Rob Witoff, head of platform at Coinbase, shared the figures in an interview with Cointelegraph, saying nearly every employee now uses AI daily. The latest estimate marks a sharp increase from February, when Coinbase said AI contributed to roughly 40% of its codebase.

Coinbase Executive Says 95%-100% of Company Code Is Now AI-Assisted, Up From 40% in February

Rob Witoff, Head of Platform at Coinbase, stated in Cointelegraph interview that 95% to 100% of Coinbase's code is currently written by or with the assistance of AI and large language… pic.twitter.com/tj4XSEdfCi

— Wu Blockchain (@WuBlockchain) July 15, 2026

Coinbase expands AI-assisted coding across engineering

The figure does not mean AI independently writes every line of production code. Witoff described the process as code written “by or with” large language models, with engineers remaining responsible for review, testing, and deployment.

He said AI usage depends on the task. Internal prototyping is now almost fully automated, while areas such as cryptography continue to receive close human scrutiny because they protect trading infrastructure and customer assets.

Witoff added that most Coinbase engineers now work with five to 10 AI agents at the same time. Combined, those agents perform coding work that the company estimates is equivalent to about 1,200 employees.

AI strategy follows Coinbase workforce restructuring

The update comes a few months after Coinbase cut about 700 employees, or 14% of its workforce, during a broader restructuring announced in May. CEO Brian Armstrong said at the time that AI had changed how work gets done and that the company needed to return to a startup-style operating model built around smaller teams.

The latest figures show Coinbase’s AI strategy is now embedded in its engineering workflow. Smaller engineering teams now use AI across coding, testing, debugging, and prototyping, while experienced developers continue to oversee security-sensitive systems.

The company also projects that AI agents could eventually perform work equivalent to 100,000 employees by 2030, though that figure represents a long-term target rather than a current capability.

The rapid pace of adoption also carries limits. Coinbase continues to rely on human expertise for security-critical code, reflecting the higher standard required for software that supports digital asset custody, blockchain infrastructure, and financial transactions. Coinbase’s approach also shows that faster software development has not eliminated the need for human review in security-critical systems.