Hi HN. OpenAI's internal models have written 719 math papers claiming 372 results, and almost nobody outside the field can read them. Closed Form explains each result in plain English: what was proved, what it doesn't prove, where the problem came from, and a blunt hype check. Every entry has a trust label. 172 papers list a formal Lean proof; the rest are claimed, not machine-checked, and the site never says "solved" when it only means "claimed". The original paper's wording sits next to each plain-English claim so you can check it, and every explanation has inline citations: tap one and the full source sentence it rests on opens in place. 128 of the 372 are explained so far. I'm 19 and built this with AI coding tools. It's all on GitHub, so corrections are very welcome: https://github.com/nishachay/closed-form