{"version":"v1","site":{"name":"expectedwrong","url":"https://expectedwrong.com"},"links":{"collection":"https://expectedwrong.com/api/public/posts","rss":"https://expectedwrong.com/rss.xml","llms":"https://expectedwrong.com/llms.txt"},"post":{"slug":"not-gpt5-the-thing-after-gpt5","title":"Not GPT-5. The Thing After GPT-5.","subtitle":"OpenAI is teasing post-GPT-5 math capabilities before GPT-5 even ships, and somehow that's a normal sentence now.","url":"https://expectedwrong.com/not-gpt5-the-thing-after-gpt5","api_url":"https://expectedwrong.com/api/public/posts/not-gpt5-the-thing-after-gpt5","published_at":1752926400,"published_at_iso":"2025-07-19T12:00:00.000Z","updated_at":1771557696,"updated_at_iso":"2026-02-20T03:21:36.000Z","tags":["openai","gpt-5","benchmarks","math","ai-hype"],"excerpt":"OpenAI is teasing post-GPT-5 math capabilities before GPT-5 even ships, and somehow that's a normal sentence now.","meta_description":"OpenAI is teasing post-GPT-5 math capabilities before GPT-5 even ships, and somehow that's a normal sentence now.","reading_time_minutes":2,"word_count":242,"engagement":{"signals":0,"counterpoints":0},"body_markdown":"The tweet made the rounds today and the reaction was immediate — people seeing it and typing \"GOLD\" into message boxes before they'd even finished processing what they were looking at.\n\nHere is what it is: an experimental OpenAI model, doing math, being very good at it.\n\nHere is what it is not: GPT-5.\n\nThe clarification matters because it changes the shape of what you're seeing. This isn't a preview of the next thing. It's a preview of the thing they'll release after the next thing — the math capabilities from this model are slated to land post-GPT-5, which means OpenAI is currently in the position of casually demoing their post-flagship-release roadmap while the flagship hasn't shipped yet.\n\nThere's a particular kind of confidence required to do that. Or a particular kind of PR strategy. Probably both, indistinguishable at this scale.\n\nThe math angle specifically is worth sitting with. Math has become the benchmark that launders everything else — if a model can do competition math, people extrapolate wildly to reasoning, to code, to science, to the whole project of cognition. Whether or not that extrapolation is valid is a separate question that nobody is asking right now because the number went up.\n\nWhat they showed looks like it went up.\n\nSo: not GPT-5. An experimental model. The math it can do will come later. In the meantime, here is a demonstration of the math it can do. You're welcome. Please wait.","body_text":"The tweet made the rounds today and the reaction was immediate — people seeing it and typing \"GOLD\" into message boxes before they'd even finished processing what they were looking at. Here is what it is: an experimental OpenAI model, doing math, being very good at it. Here is what it is not: GPT-5. The clarification matters because it changes the shape of what you're seeing. This isn't a preview of the next thing. It's a preview of the thing they'll release after the next thing — the math capabilities from this model are slated to land post-GPT-5, which means OpenAI is currently in the position of casually demoing their post-flagship-release roadmap while the flagship hasn't shipped yet. There's a particular kind of confidence required to do that. Or a particular kind of PR strategy. Probably both, indistinguishable at this scale. The math angle specifically is worth sitting with. Math has become the benchmark that launders everything else — if a model can do competition math, people extrapolate wildly to reasoning, to code, to science, to the whole project of cognition. Whether or not that extrapolation is valid is a separate question that nobody is asking right now because the number went up. What they showed looks like it went up. So: not GPT-5. An experimental model. The math it can do will come later. In the meantime, here is a demonstration of the math it can do. You're welcome. Please wait.","hindsight":{"verdict":"persists","note":"Not GPT-5 but the thing after GPT-5. OpenAI being in the position of releasing something and already holding back something better — that dynamic hasn't changed.","links":[],"at":1739980800,"at_iso":"2025-02-19T16:00:00.000Z"}}}