Between 95% and 100% of code at Coinbase is now written or assisted by AI, up from roughly 40% in February. The speed of that shift - disclosed by Coinbase's Head of Platform, Rob Witoff, in a Cointelegraph interview - signals a fundamental reworking of how software engineering is done inside a major regulated fintech.
Daily AI adoption across the organization
Nearly all Coinbase employees use AI tools daily, according to Witoff. That level of saturation goes well beyond a pilot program or a single engineering team. It reflects a company-wide change in how work gets done, not just a tool upgrade.
For many large companies, daily AI use remains fragmented. Coinbase compressed what often takes years into a single product cycle. The engineering team's workflow offers a sharper view of what that looks like in practice.
How engineers run AI agents
Engineers at Coinbase typically run between 5 and 10 AI agents simultaneously while working on coding tasks. These are not passive autocomplete suggestions. They are active agents handling work in parallel, allowing developers to oversee multiple streams of development at once.
The multiplier effect changes the human-to-output ratio. A single engineer managing a fleet of agents produces more than one person coding alone. That shift is central to the productivity numbers Witoff described.
The productivity equivalent of 1,200 employees
Witoff put a concrete figure on the cumulative output: the AI coding work at Coinbase now equals roughly 1,200 human employees. For a company that has publicly navigated cost pressures and headcount decisions, that number carries real strategic weight.
It also reframes the conversation about AI and jobs. Coinbase's framing is less about displacement and more about what AI allows the existing workforce to achieve. The productivity math is hard to ignore, even if the longer-term implications remain up for debate.
Projecting to 100,000 employees by 2030
If current trends hold, Coinbase expects AI agents to handle work equivalent to 100,000 employees by 2030 - an eighty-fold increase from today's 1,200-employee equivalent. Witoff shared that projection as a directional signal, not a firm forecast. Even so, it reveals how Coinbase's leadership views AI: not as a supplement to human engineering, but as its primary engine over time.
For developers tracking the impact of AI on coding, the speed and scale of the shift at Coinbase offer a concrete example of what's possible when AI is woven into the engineering pipeline from the start. Resources like AI for IT & Development can help professionals understand the tools and practices driving this kind of change.
Implications for fintech engineering
The broader takeaway from Coinbase's AI push is not just about one company's efficiency gains. A platform that handles real user funds, real transactions, and real regulatory requirements moving to near-total AI-assisted code development sets a precedent. Rivals in crypto and fintech will need to answer with their own strategies - or risk falling behind on speed, cost, and scale.
Why this matters for IT and development professionals
Coinbase's numbers - from 40% AI-assisted code to 95-100% in months, and the daily use of multiple AI agents per engineer - show that the shift from AI experimentation to full-scale engineering integration is happening faster than many expected. For developers and IT leaders, the implication is clear: the working model of a software engineer is changing, and the gap between companies that treat AI as core infrastructure and those that treat it as optional is widening quickly.
Your membership also unlocks:
