Jacob Coxon, who spent 3 года on model pretraining at OpenAI and Anthropic, publicly resigned and left the industry. He wrote on X that neither lab acts responsibly, asserting both pursue self‑improving superintelligence and thereby risk public safety.

Обвинения и внутренний язык

Coxon reported that within research teams employees use terms such as "crunchtime" and "endgame" when discussing product roadmaps and deployment schedules. He warned his projection that by the end of next year the situation could exceed internal control mechanisms and oversight.

Признаны проблемы выравнивания

The resignation followed a recent admission by OpenAI chief scientist Jakub Pachocki that no laboratory has solved alignment at the scale now being pursued. Coxon said Anthropic recognises the dilemma publicly but faces competitive pressure to move forward.

Коммерческий контекст и сроки

Observers note the timing coincides with Anthropic preparing for an IPO potentially valued at $2 T, intensifying incentives to release more capable models ahead of public listing. Coxon emphasised that his perspective comes from engineering roles rather than activist circles.

Сообщаемые убеждения и последствия

According to Coxon, some developers genuinely believe that advanced systems could pose existential risks before the close of the decade. He framed his departure as a response to perceived lack of sufficient restraint across competing organizations.