David Robinson, who used to write the safety reports accompanying every major model release at OpenAI, resigned this week and has taken his concerns public in an editorial in The Atlantic. Yes, it's another one of these. No, you're not wrong to feel a little cynical about people warning us about the very thing they helped build. But the warnings, unfortunately, remain worth reading.

Robinson's argument is that the problem isn't a missing rule or a lax regulation here and there - it's the culture itself. Silicon Valley, he says, runs on "extreme confidence" and "perpetual sprints," churning out bigger and better models with "unimpeded optimism" while ignoring or underestimating the potential problems those models bring with them.

His proposed fix is not subtle. AI companies, he argues, need humility and a willingness to look beyond the insular, move-fast-and-break-things world of tech. Specifically, he says frontier labs should operate less like startups and more like nuclear power plants or busy airports - with layers of redundancy and careful, time-consuming planning, so that the occasional and inevitable human error doesn't open a door to disaster.

Robinson is the latest in a lengthening parade of researchers and safety workers walking away from prominent AI firms. Jacob Coxon appears to have kicked off the exodus by quitting Anthropic and then publicly stating that AI "could kill us all by the end of the decade." He was followed by Robert O'Callahan, Bilal Chughtai, and Josh Engels at Google DeepMind, along with Joe Benton at Anthropic.