OpenAI job listing tracks 'automation of technical staff' amid self-improving AI bets

imjustnewatai · x · 2026-09-05

An OpenAI job listing literally says "track progress toward automation of technical staff," fueling speculation about the post-Astra roadmap. The author bets the next model gets a research goal and a compute budget: remembering failed experiments, dividing work across agents, writing and running code, checking results and deciding what to try next. The gpt-next training role explicitly includes memory and multi-agent capabilities. Astra is reportedly already tested on improving small models with one H100 in five hours, with the agent choosing its own training data, methods and follow-up experiments. The next leap: reliable multi-day work reproducing papers, fixing broken runs and delivering improvements — then using that system to develop the generation after it.

Original post →

More from AGI Musings

AGI Musings channel →