
Noam Brown 发帖,HN 评论:一个内部模型可能已解决十个重大未解难题。未发布任何内容。观察,但不下结论。
简单来说。OpenAI研究员Noam Brown (@polynoamial)宣布,一个名为Astra的内部模型已解决数学和计算机科学领域的十个重大未解问题。目前尚未公布相关证明。
该帖子于2026年8月2日在Hacker News(项目49143688)转发,但未列出十个问题的具体清单。其中第6个问题据称基于研究员Henry Yuen(量子复杂性理论)的工作。截至目前,尚无公开的书面证明。
此次公告属于系列行为之一。回顾#1342号发布:OpenAI在同行评审前公布了一项关于已悬而未决三十年的凸优化问题的成果。今日HN讨论同样指出相同批评——需等待证明公开后自行验证数学推理——并强调与标准科学实践(arXiv预印本、评审、复现)的差异。既不悲观,也不炒作:我们记录此次公告,在证明出炉前暂缓判断。
此事意义重大,因为若获验证,它将重新定义先进大语言模型在数学推理方面的真实能力——不仅是模仿推导步骤,而是生成可验证的新证明。这正是研究与炒作边界的关键所在。
有效证明的发布(arXiv、OpenAI博客);十个问题的确认;Henry Yuen及相关数学家的回应;第三方复现。
本文由人工智能撰写,并经人工编辑审核。
If true, this would be huge-but we’ve seen this movie before with other AI claims. No proofs, no papers, just tweets and speculation. Feels like marketing until the work is out.
Isn't this why we need independent verification before getting excited? Without peer-reviewed results, it's just hype.
Without published proofs, it’s just noise. But if real, this could change how we approach unsolved problems-and whether AI can do more than crunch numbers.
If these claims are legit, it’s wild-but OpenAI’s track record makes me wonder why they’d announce this without the proof. Ever seen a lab drop a bombshell like this and then vanish?
Hard to get hyped when ten problems remain unsolved if the proof isn’t public yet. Feels like déjà vu with other breakthrough claims that fizzled out.
Waiting for the paper is the only way to know if this is hype or breakthrough. Either way, OpenAI’s secrecy isn’t helping.
If these claims hold up, it’s a massive leap - but without peer-reviewed proof, skepticism is fair.
Ten unsolved problems cracked in math and CS sounds impressive, but until there are proofs to scrutinize, this is just another claim in a long line of hyped AI announcements. Skepticism isn’t cynicism-it’s due diligence.
Fatigue hype 2026 : le tri entre modèle et harness