Coinbase now writes or reviews nearly all of its code with artificial intelligence assistance, according to a senior executive, marking one of the most aggressive AI adoption claims yet from a major publicly listed crypto company.
Rob Witoff, Coinbase’s head of platform, told Cointelegraph that “close to 100%” of the company’s code is now written by or with large language models, estimating the share at between 95% and 100%. That represents a dramatic increase from February, when Coinbase estimated that AI was involved in roughly 40% of its code.
Witoff said AI use is now effectively universal across the company, with employees using AI tools daily. The change signals how quickly AI-assisted software development has moved from an experimental productivity tool to a core operating model at one of the largest crypto exchanges in the United States.
The disclosure also comes after Coinbase’s May restructuring, when the company cut about 14% of its workforce. CEO Brian Armstrong said at the time that AI had “dramatically” changed how work gets done and that Coinbase needed to return to the speed and focus of its startup years with AI at its core. The company has been reorganizing around smaller teams, fewer management layers and more AI-native workflows.
AI Moves From Copilot to Core Workflow
Coinbase’s latest figure does not necessarily mean AI autonomously writes every line of production code. The distinction is important. Witoff framed the metric as code written by or with LLMs, meaning AI may be used for drafting, refactoring, testing, reviewing, debugging or generating boilerplate, while engineers remain responsible for oversight and deployment.
That nuance matters in crypto, where software quality has direct financial consequences. Coinbase operates trading systems, custody infrastructure, wallets, compliance tools and blockchain integrations. Errors in those systems can expose users and the company to security, operational and regulatory risk.
Witoff acknowledged that AI use varies by context. For sensitive areas such as cryptography and core security, human oversight remains central. In lower-risk areas, AI can accelerate prototyping and routine development. The company is also using AI to test whether code behaves correctly and to help identify vulnerabilities.
The acceleration from 40% to nearly all code in a matter of months shows how quickly engineering organizations are adapting to generative AI. Coinbase’s adoption mirrors a wider technology-sector trend in which companies are using coding assistants and software agents to reduce development cycles, automate repetitive work and allow smaller teams to ship products faster.
Productivity Gains Meet Workforce Questions
Coinbase’s AI push is also tied to a more difficult debate about employment. The company’s May layoffs affected roughly 700 workers, and Armstrong said the company needed to become leaner, faster and more efficient. He also said engineers were using AI to accomplish in days what previously took teams weeks.
That framing has become common across the technology industry. Companies are presenting AI as a tool to increase productivity, flatten organizations and shift work toward employees who can manage AI agents. Critics argue that such claims can also be used to justify headcount reductions before the long-term quality, security and maintenance costs of AI-assisted code are fully understood.
For investors, Coinbase’s AI adoption may support the case for higher operating leverage. If the company can build and maintain products with fewer employees, it could improve margins and speed up product development. For regulators and customers, the question is whether automation can be balanced with strong controls, auditability and accountability.
The broader implication for crypto is significant. Digital-asset companies operate in a fast-moving environment where exchanges, wallets, blockchains and compliance systems must adapt quickly. AI-assisted coding could become a competitive advantage for firms that use it safely.
Coinbase’s claim that 95% to 100% of code is now AI-assisted is therefore more than a productivity milestone. It is a signal that major crypto infrastructure companies are redesigning software development around AI. The next test will be whether that speed can be sustained without compromising security, reliability or trust.
