What now ...?
retained text
so ... now we have to be nice to a gradient descent? It's not even AI, it's an LLM. fun fact: research\* showed that being ruder or more direct to models can increase accuracy compared to being polite. \*Research paper: [https://arxiv.org/pdf/2510.04950](https://arxiv.org/pdf/2510.04950)
document fetched 2026-10-10 00:40 · canonical https://reddit.com/r/ClaudeCode/comments/1x18ap0/what_now