# Stop LLMs from over-engineering skills and rules with the bitter lesson skill — retained record

> Retained reddit record from reddit:ClaudeAI.

Canonical: https://highsignal.app/data/records/11000f398134b943

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:ClaudeAI)  published 2026-10-23 20:44 ingested 2026-10-24 10:56

[original source ↗](https://www.reddit.com/r/ClaudeAI/comments/1xdpj70/stop_llms_from_overengineering_skills_and_rules/)

## retained text



```
LLM love to create overengineered rules that they themselves are too smart to actually need any more. If you just remind it about the bitter lesson, then they will go 'Oh yeah! That's right!' and stop doing it.

The original 'bitter lesson' was related to the first chess computers. Basically rather than training the chess computer on all of human knowledge of chess, it was better to just make it play against is self millions and millions of times.

[https://en.wikipedia.org/wiki/Bitter\_lesson](https://en.wikipedia.org/wiki/Bitter_lesson)

    ---
    name: consider-the-bitter-lesson
    description: Consider the bitter lesson in the current task.
    disable-model-invocation: true
    ---
    
    Consider the bitter lesson (Richard Sutton, 2019) when this skill is invoked.
```



document fetched 2026-10-24 10:24 · canonical https://reddit.com/r/ClaudeAI/comments/1xdpj70/stop_llms_from_overengineering_skills_and_rules
