OpenAI developer Thibault Sottiaux poured some gas on the hype fire. Astra was probably OpenAI's "biggest competitive advantage" while it wasn't publicly available, he wrote on X. Since the team started using it internally, productivity jumped so much that some plans got moved up by six months.

That tracks with a recent study by IAPS fellow Severin Field. Twenty out of 25 researchers surveyed at OpenAI, Anthropic, Google DeepMind, and Meta ranked the automation of AI research as one of the biggest AI risks. Several milestones they flagged have already been hit. Half the respondents expect the most powerful models to stay internal and never be sold to the public. Anthropic claims, in a similar vein, that Claude now writes over 80 percent of its own production code. But not everyone buys the productivity boost, especially when it comes to AI models improving themselves.