Kelsey Tuoc: OpenAI and Anthropic staff believe it'll be bad if the other gets recursive AI first

EigenGender · x · 2026-09-10

Kelsey Tuoc says people at both OpenAI and Anthropic believe that recursively self-improving AI, if they get it right, would be the best thing that ever happened — but bad if the other lab gets there first. This mindset, she argues, makes them reluctant to slow down even as evidence accumulates that nobody knows how to do it right. EigenGender adds the situation is importantly asymmetric; if it were symmetric, it would be a clearer coordination failure.

Related event: AI Labs All Believe They Can Win the Race, So No One Brakes(3 posts)→

Original post →

More from AGI Musings

AGI Musings channel →