Cryptocurrency exchange Coinbase has taken software code creation to a new level of automation. According to internal data, over 95% of the platform's code is now generated using artificial intelligence. Coinbase CEO Rob Witoff noted that this figure has doubled compared to February, when AI's share in development stood at 40%. Today, neural networks are involved in the daily work of 100% of the exchange's employees.
The degree of automation varies depending on the complexity of tasks. Prototyping is fully automated, core systems operate in a hybrid mode, and the critically important area—cryptography—remains under strict developer control. In this field, AI acts as an assistant: it helps identify vulnerabilities and verify mathematical calculations, but every line of code undergoes manual review.
This approach has already led to significant changes in team structure. In May, Coinbase cut 700 employees (14% of its workforce). CEO Brian Armstrong explained this as a need to restore "startup speed and focus" through AI. Now, groups of 2–3 senior specialists handle tasks that previously required the involvement of 10 or more people. The cuts primarily affected junior developers, as well as employees in marketing, support, and compliance.
The scale of implementation is impressive: each Coinbase engineer has between 5 and 10 active AI agents. Collectively, they perform work equivalent to that of 1,200 employees. The company predicts that by 2030, this figure will grow to 100,000 virtual workers.
This trend is not unique to Coinbase. In April, Snap replaced 1,000 employees with neural networks to save $500 million, and in May, analytics platform Dune announced a 25% staff reduction after restructuring. Coinbase has also launched a service for AI agents, confirming its strategic focus on automation.
My analysis: Coinbase demonstrates how AI is transforming not only development but also the business model of crypto companies. Reducing the share of manual code work to 5% is a revolution, but it carries risks: reliance on AI could make systems vulnerable to new types of attacks. Nevertheless, for investors, this signals increased efficiency and reduced operational costs in the long term.