OpenAI Group PBC revealed on August 2 that an internal version of Astra, the model family it describes as its next major release, produced new results for 10 problems in mathematics and theoretical computer science that had been open for at least a decade2.
The company published machine-checkable proofs alongside the claim. OpenAI posted a 249-page manuscript detailing the results.
The announcement drew attention on Hacker News, where the thread accumulated dozens of comments1.
ANALYSIS The inclusion of machine-checkable proofs is a notable methodological choice: it subjects the results to formal verification rather than relying solely on peer review of natural-language arguments. If the proofs withstand scrutiny, the work would represent a concrete demonstration of an AI system contributing original results to open research problems in pure mathematics and theoretical computer science, rather than reproducing known solutions or performing well on curated benchmarks.