AI Now Assists in Writing Over 95% of Coinbase’s Code, Driving Shift to Leaner Teams

The intersection of artificial intelligence and Web3 is moving far beyond theoretical use cases, reshaping the very infrastructure of the digital asset economy. In a striking revelation of how deeply automation has penetrated the upper echelons of the industry, Coinbase has disclosed that between 95% and 100% of its software code is now written by or with the assistance of large language models (LLMs).

This rapid technological pivot follows a significant corporate restructuring in May, during which the leading American crypto exchange reduced its workforce by 14%, parting ways with approximately 700 employees. At the time, Coinbase CEO Brian Armstrong framed the layoffs not merely as a cost-cutting measure, but as a strategic realignment. In an email to staff, Armstrong explained that AI had “dramatically” accelerated the pace of development, necessitating a “return to the speed and focus of our startup founding, with AI at our core.”

The LLM Takeover of the Codebase

The speed of this transition has caught many industry observers by surprise. Just months earlier, in February, Coinbase estimated that roughly 40% of its code was being generated with the help of AI. The leap to near-total integration highlights how aggressively the firm has deployed automated development tools.

“Effectively, 100% of our employees are using AI on a daily basis here,” said Rob Witoff, Coinbase’s head of platform, in an interview with Cointelegraph. “And close to 100% of our code, probably somewhere between 95% and 100%, is written by or with LLMs today.”

However, Witoff emphasized that this does not mean human engineers have been entirely replaced by autonomous algorithms. Instead, there is a “wide spectrum” of implementation across the company’s technology stack.

For high-stakes security components, such as core cryptography, the process remains heavily reliant on human expertise. Conversely, the creation of internal prototypes has become entirely automated. Core transactional and operational systems occupy the middle ground of this spectrum.

“We’re using AI quite a bit to test and make sure the code we’ve written is working the way it should, there’s no vulnerabilities, we’re verifying the math, but that’s a much more manual part than where we’re building internal prototypes, which is now effectively a 100% automated,” Witoff explained.

This hybrid approach is critical in the digital asset space, where smart contract vulnerabilities or cryptographic flaws can lead to catastrophic, irreversible financial losses. By automating the tedious aspects of prototyping and initial drafting while retaining rigorous human oversight for mathematical verification and security auditing, Coinbase aims to balance rapid innovation with institutional-grade security.

Flatter Teams and the Rise of AI agents

This technological evolution is fundamentally altering the organizational structure of tech companies. Rather than maintaining large hierarchical engineering departments, Coinbase is reorganizing around smaller, highly specialized teams led by senior “tastemakers.” According to Witoff, teams of just two or three senior engineers are now capable of executing projects that previously required ten or more people.

This shift has inevitably altered the hiring landscape within the blockchain sector. “There were a lot of junior development roles that were impacted,” Witoff noted regarding the May layoffs. However, the workforce reductions were not isolated to engineering; they swept across various departments, including marketing, legal, customer support, and compliance.

To maintain its operational output with a leaner staff, Coinbase has leaned heavily on autonomous AI agents. Witoff revealed that the majority of Coinbase engineers now manage between five and ten AI agents operating simultaneously. Collectively, these automated agents are performing coding tasks equivalent to the output of approximately 1,200 full-time human employees.

Looking ahead, Coinbase projects an exponential increase in this automated capacity. By 2030, Witoff estimates that AI agents could perform work equivalent to a staggering 100,000 human employees, all while keeping the company’s direct overhead remarkably lean. Notably, Coinbase has reported that its overall AI expenditures have remained “flat” despite the escalating volume of token transactions and computational demands generated by these systems.

A Broader Paradigm Shift in Web3

Coinbase is far from alone in its aggressive embrace of AI-driven operational efficiency. The broader cryptocurrency and fintech sectors have experienced a wave of restructuring as executives leverage new automation tools to streamline operations.

In March, rival crypto exchange Crypto.com reduced its headcount by 12%, specifically targeting roles “that do not adapt in our new world.” Similarly, Block CEO Jack Dorsey announced a sweeping 40% reduction in his company’s workforce in February, explicitly tying the decision to the leverage provided by modern artificial intelligence.

“We’re already seeing that the intelligence tools we’re creating and using, paired with smaller and flatter teams, are enabling a new way of working which fundamentally changes what it means to build and run a company,” Dorsey shared in a post on X.

Other prominent firms in the digital asset ecosystem—including Kraken, Gemini, Messari, and Dune—have also cited AI-related efficiencies when executing workforce reductions this year. Meanwhile, traditional retail brokerage platforms are looking to bring these capabilities directly to users. Robinhood, for instance, recently indicated that its own proprietary AI agent feature will “soon” be assisting crypto traders with market analysis and execution.

As the line between software engineering and AI orchestration continues to blur, the digital asset industry is serving as a real-world testing ground for the future of work. For companies like Coinbase, the goal is clear: leverage autonomous agents to scale operational capacity to unprecedented heights, proving that in the Web3 era, a lean, AI-empowered team can outpace the legacy giants of finance.