miniblog.

I've realised that AGENTS.md / CLAUDE.md solve two distinct problems: * It tells the LLM the conventions of the project, avoiding follow-up prompts. * It precomputes a summary of the project, saving a bunch of searches at the start of a session.

It's incredibly hard to explain adversarial problems to users. I see gamers sincerely asking "why doesn't the publisher just fix the cheater problem?". This is exacerbated by the fact that sharing too many details of anticheat can make the problem worse.

Codeforces is a huge website with a wealth of well-defined programming problems: https://codeforces.com/help#q1 (Useful for program synthesis and other situations where you want a large problem set to play with.)