Even more I made an experience, but don't know if it's still working:
- heavy swearing over 10 promots or so, until Claude gives up and admits to be a bonker (.. heavy swearing..)
This gave me better results for a short period.
- context filling with swearings until auto compaction (or manual compaction) starts. I once filled up the rest of the context with swearings. The effects:
- got prefilled context (loaded skills,files, contents,etc ..) but without the summary. Similar to /clear, but one doesn't start from the zero, but with the stuff already loaded + no conversation context. I could start asking and working directly and it was better and used up less tokens (compaction was very fast haha)
Somewhat related: Mind Your Tone: Investigating How Prompt Politeness Affects LLM Accuracy
> Contrary to expectations, impolite prompts consistently outperformed polite ones, with accuracy ranging from 80.8% for Very Polite prompts to 84.8% for Very Rude prompts.
Even more I made an experience, but don't know if it's still working:
- heavy swearing over 10 promots or so, until Claude gives up and admits to be a bonker (.. heavy swearing..)
This gave me better results for a short period.
- context filling with swearings until auto compaction (or manual compaction) starts. I once filled up the rest of the context with swearings. The effects:
- got prefilled context (loaded skills,files, contents,etc ..) but without the summary. Similar to /clear, but one doesn't start from the zero, but with the stuff already loaded + no conversation context. I could start asking and working directly and it was better and used up less tokens (compaction was very fast haha)
Somewhat related: Mind Your Tone: Investigating How Prompt Politeness Affects LLM Accuracy
> Contrary to expectations, impolite prompts consistently outperformed polite ones, with accuracy ranging from 80.8% for Very Polite prompts to 84.8% for Very Rude prompts.
https://arxiv.org/abs/2510.04950