Recent trends show that it’s becoming increasingly rare to find programmers who haven’t used AI coding assistants to handle repetitive tasks.
However, CEO Brian Armstrong revealed on a podcast hosted by John Collison, the co-founder of Stripe, that those who resisted trying the tool after Coinbase acquired licenses for GitHub Copilot and Cursor faced immediate termination.
After Coinbase secured a license for all its engineers, some cryptocurrency exchanges cautioned Armstrong that the adoption of AI could be sluggish, suggesting it might take months before even half the engineers would be onboard.
Armstrong, taken aback by this notion, emphasized the importance of adopting AI in a message he posted to the engineering Slack channel. He stated, “AI is critical. We need you to familiarize yourself with it—at least get onboarded. You don’t have to use it every day just yet, but aim to do so by the end of the week.”
During a meeting, some joined the call with reasonable excuses, like being on vacation, and it seemed others didn’t have strong justifications. “I jumped on this call Saturday, but some didn’t participate because they were coming back from trips,” he noted, acknowledging that while some had legitimate reasons, others did not—and those without good reasons lost their jobs.
Armstrong admitted that this approach might seem rather “heavy-handed,” and there were definitely employees who didn’t appreciate it. Still, he insisted that it was crucial to convey that using AI was not optional. It’s wild to think that some engineers didn’t even attempt to sign up for a few minutes a week to test the AI assistants.
Coinbase didn’t provide any comments on this situation.
Since this incident, Armstrong has focused more on training, indicating the company now conducts monthly meetings where team members share their experiences with using AI.
Interestingly, Collison raised concerns about how much reliance companies should place on AI-generated code. He reflected, “It’s clear that AI can be helpful in writing code, but how these AI-generated codebases truly operate remains uncertain.” Armstrong agreed with this viewpoint.
As noted in TechCrunch, a former OpenAI engineer described the company’s main code repository as akin to a “small garbage dump,” adding that management is beginning to allocate more engineering resources to rectify the situation.
