Ask HN: Is AI good enough for math exercises?

A user asks whether AI is reliable enough to correct solutions to undergraduate math exercises from books such as Tao's Analysis and Spivak, rather than to teach the material. They have access to Gemini Pro and a MacBook M5 that might run a Qwen model locally, and ask whether to expect reasoning mistakes or hallucinations, noting they are not strong at proofs and want to improve.