Coinbase CEO Revealed He Fired Engineers Who Didn't Use AI Tools

Coinbase CEO Brian Armstrong fired engineers in 2025 after giving them one week to adopt AI coding tools — then holding a Saturday meeting where those without a valid excuse were terminated on the call. Armstrong disclosed the story publicly in a conversation with Stripe President John Collison, confirming that 33% of Coinbase's code is now written by AI.

The story has become a flashpoint in the tech industry's debate over AI adoption. Armstrong's framing — that resisting AI tools is a performance issue, not a personal preference — signals a shift in how tech leadership views developer workflow. Engineers preparing for roles at AI-forward companies should expect interview questions about their hands-on experience with tools like Cursor; practice those answers with an AI mock interview before your next technical screen.

Why Did Coinbase CEO Fire Engineers Who Refused AI Tools?

Armstrong gave engineers one week to adopt AI tools after his team proposed a slow, quarter-long rollout — and he rejected it outright. The terminations happened on a Saturday call Armstrong convened specifically for the holdouts.

When his team proposed a gradual adoption plan, Armstrong pushed back directly: "Originally they were coming back and saying, 'All right, over the next quarter... we're going to get to 50% adoption.' I said, 'You're telling me — why can't every engineer just onboard by the end of the week?'"

He posted in the company's all-hands Slack channel with a clear ultimatum: "AI's important. We need you to all learn it and at least onboard. You don't have to use it every day yet until we do some training, but at least onboard by the end of the week. And if not, I'm hosting a meeting on Saturday with everybody who hasn't done it and I'd like to meet with you to understand why."

The Saturday call produced immediate consequences: "I jumped on this call on Saturday and there was a couple people that had not done it. Some of them had a good reason, because they were just getting back from some trip or something, and some of them didn't and they got fired." Armstrong acknowledged the backlash: "Some people really didn't like it, by the way, that heavy-handed approach but I think it did set some clarity at least that we need to lean into this."

How Much of Coinbase's Code Is Written by AI?

The mandate delivered measurable results. Armstrong confirmed that 33% of Coinbase's code is now generated by AI tools, with a stated goal of reaching 50% AI-written code by Q3 2025. As of 2026, Coinbase has continued its AI-first engineering culture, running mandatory code reviews on all AI-generated output. That pace — moving from near-zero to one-third of all code in under a year — is among the fastest documented AI adoption rates at a major financial technology company.

To sustain code quality at that scale, Coinbase runs monthly "AI Speed Runs" — sessions where a top-performing engineer demonstrates their most effective AI tool techniques to the rest of the team. The program identifies the engineers using AI most effectively and spreads those practices company-wide.

Vibe coding is the practice of using AI prompts to generate entire features or functions without the developer deeply understanding or reviewing the underlying code. Armstrong's concern is that engineers who vibe code financial logic introduce bugs that pass automated tests but fail in edge cases involving real money. Browse more tech industry news on how AI mandates are reshaping engineering teams.

Armstrong was direct about the limits of AI-generated code in financial systems: "You don't want people vibe coding these systems moving money. We've really encouraged people to really, you have to code review it and have the appropriate checks in place on that from with humans in the loop."

Did Coinbase's AI Mandate Apply to Non-Engineering Teams?

Armstrong's directive extended far beyond the engineering org. He described pushing AI adoption across design, product management, and finance functions simultaneously.

"We want to make sure it's used not just in the engineering teams. It really should be any team — design is using it heavily. Product managers, I think FP&A could even be using this to generate, ingest all the data and tell me what you forecast the revenue to be," Armstrong explained.

The AI integration has reached executive decision-making. Coinbase's RAPIDS framework — the company's internal decision-making protocol — now includes an AI participant that contributes analysis alongside human reviewers. Armstrong said the company is actively testing how much decision authority AI can hold versus humans, including in some areas where AI may already outperform individual judgment. For engineers pivoting to product or cross-functional roles at AI-first companies, an AI resume builder can help frame your AI tool experience in a way that resonates with these hiring managers.

What Should Software Engineers Do About AI Tool Mandates in 2026?

The Coinbase situation reveals a practical contradiction at the center of the AI adoption push: Armstrong wants 50% of Coinbase's code written by AI but explicitly prohibits vibe coding. He is mandating AI tool usage while requiring human code review for every AI-generated line. The tools are useful enough to demand adoption and unreliable enough to require constant oversight — simultaneously.

For engineers, this means the skill being tested is not whether you use AI, but whether you can use it without losing your engineering judgment. The engineers at Coinbase who were fired weren't wrong to be cautious about AI-generated code in financial systems — they were wrong to refuse learning the tools entirely.

Armstrong's warning about vibe coding destroying junior developer careers reflects a concern shared across senior engineering leadership: that AI tools, used poorly, produce engineers who can prompt but cannot debug. The engineers who thrive in AI-forward companies treat AI as an accelerant for their existing skills, not a replacement for deep understanding.

Three steps for engineers navigating this shift:

• Adopt the tools before forming opinions. Armstrong's firings weren't about quality concerns — they were about refusal to engage. Onboard with Cursor or GitHub Copilot, then evaluate.
• Build a code review discipline. Armstrong's mandatory review requirement is the right instinct. Develop a personal checklist: logic correctness, edge cases, security implications, test coverage.
• Practice articulating your AI workflow in interviews. Technical interviews at AI-forward companies increasingly include questions about your specific AI tool approach. Use Interview Copilot to sharpen those answers before your next screen.

Related Blogs

• AI Job Displacement 2025: Which Jobs Are At Risk? — How AI adoption mandates like Coinbase's are reshaping employment across tech, with data on which engineering roles face the most pressure.
• Computer Science Graduates Face the Worst Job Market in Decades — What collapsing entry-level tech hiring means for engineers entering a market where AI tool fluency is now table stakes.
• 30 Jobs Most Impacted by AI in 2025 — Indeed's ranking of which occupations are seeing the greatest AI-driven skill transformation, including software development at 81%.
• Software Engineering Job Market 2026 — The current state of software engineering hiring and how AI mandates like Coinbase's are changing what employers look for in 2026.

Frequently Asked Questions

Did Coinbase CEO actually fire engineers for not using AI tools?

Yes. Coinbase CEO Brian Armstrong confirmed he fired engineers who refused to adopt AI coding tools in 2025. He gave the team one week to comply, held a Saturday meeting for holdouts, and terminated those without a valid reason on the call.

What AI tools did Coinbase require engineers to use?

Coinbase required engineers to adopt Cursor and GitHub Copilot as their primary AI coding tools. Armstrong set a one-week deadline for all engineers to onboard with both tools.

How much of Coinbase's code is now written by AI?

As of Armstrong's public disclosure, 33% of Coinbase's code is written by AI. The company set a Q3 2025 target of reaching 50% AI-written code, with mandatory human review for all AI-generated code.

Does Coinbase allow vibe coding?

No. Armstrong explicitly warned against vibe coding in financial systems. All AI-generated code at Coinbase requires human code review before merging, and engineers are expected to understand the code they ship regardless of whether AI produced it.

Is Coinbase still requiring AI tool usage in 2026?

Yes. As of 2026, Coinbase's AI-first engineering mandate remains in effect. The company has maintained mandatory code reviews on all AI-generated output and continues its monthly AI Speed Run training sessions to elevate team-wide AI tool fluency.

What does the Coinbase AI mandate mean for software engineers?

The Coinbase situation signals that AI tool fluency is now a hiring requirement at many tech companies. Engineers who use AI tools thoughtfully — adopting them without losing engineering judgment — have a significant advantage. Join the Final Round AI community to connect with engineers sharing strategies for navigating AI adoption at work.