Amateur armed with ChatGPT solves an Erdős problem | Hacker News

A Hacker News thread focused on an Erdős-style number theory problem about primitive sets and a proposed proof strategy using ChatGPT.

The post drew discussion on whether the argument was sound, with users weighing mathematical rigor, heuristics, and proof verification.

The thread broadened into a debate over LLM math ability, with examples of success and failure in reasoning, computation, and theorem-like tasks.

Commenters compared AI’s strengths to human mathematical intuition, noting limits such as finite context windows and unreliable proof generation.