Show HN: Mentwire — Paul Graham's 'Intellectual Captcha' Idea Built as an LLM-Powered Bot Filter

Hacker News·low signal

A developer built Paul Graham's concept of an 'intellectual captcha' — instead of image puzzles, users answer open-ended questions that test genuine understanding, with an LLM evaluating whether the response is human-quality reasoning or bot-generated. 29 points but 46 comments on HN (1.59 comment-to-point ratio), indicating intense debate about whether LLM-graded captchas create a circular dependency problem where bots that can pass are the same ones being filtered.