Noam Brown, an OpenAI research engineer, posted on X that the company devotes resources to announcing math results from internal models like Astra only when the findings would meaningfully change people’s understanding of AI progress. He added that the main focus remains shipping great models so the public can use them. The post follows an earlier announcement that an internal Astra version solved ten open problems in mathematics and theoretical computer science. It drew replies on X and was retweeted by fellow researcher Mo Bavarian.