x.com/ValerioCapraro/status/2097791836269977996
1 correction found
one of the ten problems OpenAI later announced Astra had solved.
This overstates OpenAI’s own announcement. OpenAI said Astra produced ten results that either resolved or made substantial progress on open problems, not ten fully solved problems.
Full reasoning
OpenAI’s August 1, 2026 announcement does not say Astra solved ten open problems. It says OpenAI was sharing "a selection of ten results, each of which resolves or makes substantial progress on a long-standing open problem." The list itself confirms that several items were partial advances rather than complete solutions—for example, "New upper bounds on sphere-packing density", "Exponentially improved bounds" for codes, and "New lower bounds" in arithmetic circuit complexity.
So while the non-sofic groups result may indeed be one of the ten results and may count as a full resolution of that particular question, the post’s wording incorrectly implies that OpenAI announced Astra had solved ten problems. A more accurate description is that OpenAI announced ten results, some of which were full resolutions and others of which were substantial progress.
This is also how independent reporting summarized the announcement: Scientific American described it as "10 ... math advances" and said each result "resolves or progresses" a different open problem.
2 sources
- Ten advances in mathematics and theoretical computer science | OpenAI
OpenAI wrote: "Today, we are sharing a selection of ten results, each of which resolves or makes substantial progress on a long-standing open problem." The listed results include partial advances such as "New upper bounds on sphere-packing density," "Exponentially improved bounds" for codes, and "New lower bounds" for arithmetic circuit complexity.
- OpenAI's latest math breakthroughs commit research misconduct, experts say | Scientific American
Scientific American described the announcement as "10 more artificial-intelligence-generated math advances" and wrote that "each one resolves or progresses a different 'long-standing open problem.'"