The Math Behind ‘1,200 Digital Workers’

Pointing Toward a Scary but Imminent Reality

Witoff detailed the shift in an exclusive interview published yesterday, adding:

Effectively, 100% of our employees are using AI on a daily basis here … close to 100% of our code, probably somewhere between 95% and 100%, is written by or with LLMs today.

Large language models (LLMs) are the AI systems behind tools such as ChatGPT and coding assistants, and according to Witoff, most engineers at the Nasdaq-listed exchange (Nasdaq: COIN) are now running five to ten AI agents at a time to write, test, and review software (in an effort to minimize human direction).

The pace of adoption is the striking part, given Coinbase revealed in February that 40% of its code was AI-written, meaning the share has more than doubled in roughly five months.

Witoff described a “wide spectrum” in how far the company lets the machines go. Core cryptography (the security-critical code protecting customer funds), for instance, still relies mainly on human input, while activities such as ‘prototyping’ have been fully automated. Core system management, on the other hand, sits somewhere in the middle.

That tiering system is important for an exchange holding billions of dollars in customer assets because a single flawed line can become an exploit vector overnight.

The executive also offered a forecast wherein by 2030, Coinbase could see AI agents doing the equivalent work of 100,000 employees, roughly 25 times the headcount of the company itself.

Crypto’s Broader AI Embrace

Coinbase’s numbers land amid a wider convergence of AI and crypto infrastructure as the exchange recently revealed its membership in the newly launched x402 Foundation, the Linux Foundation body standardizing AI agent payments alongside Ripple, Visa, and Mastercard.

Critics of AI-coding statistics note that “written with the help of AI” can cover everything from autocomplete suggestions to fully autonomous pull requests, making cross-company comparisons slippery. Witoff’s own 95–100% range bundles both ends of that spectrum.

What is not in dispute is the direction. An exchange that measured AI’s share of its codebase at 40% in winter now struggles to find code the machines have not touched.