OpenAI Developer Says Astra Boosted Productivity Enough to Move Plans Up Six Months

OpenAI developer Thibault Sottiaux said Astra was probably OpenAI’s “biggest competitive advantage” while it was not publicly available. According to the article, he wrote on X that internal use increased productivity enough for some plans to be moved up by six months. The claim points to Astra not just as a model or product milestone, but as a tool that may have changed OpenAI’s own development cadence.

The report connects Sottiaux’s claim to broader concern around the automation of AI research. A recent study by IAPS fellow Severin Field found that 20 out of 25 researchers surveyed at OpenAI, Anthropic, Google DeepMind, and Meta ranked automation of AI research among the biggest AI risks. If the most capable systems are first used internally, the productivity gap between frontier labs and outside users could widen before the public sees the tools.

The article notes that half of the surveyed researchers expect the most powerful models to remain internal and never be sold to the public. That makes Sottiaux’s claim especially relevant: internal-only systems could become operational advantages rather than consumer products. For readers tracking AI competition, the key signal is not just model capability, but who gets access and when.

The Decoder also notes that similar claims exist elsewhere, including Anthropic’s claim that Claude now writes over 80 percent of its own production code. At the same time, the article cautions that not everyone accepts broad productivity-boost claims, especially around AI models improving themselves. The practical takeaway: treat internal acceleration claims as important signals, but look for independent evidence before assuming they translate cleanly into measurable business or research gains.