> The system prompt used in the standalone compaction request is different. Instead of telling the LLM "you are an expert coding assistant", we tell the LLM "you are a context summarization assistant."
No disrespect to Pi or the team, it’s my daily driver, I have a lot of respect for the project. But this feels naive to me. You might as well do it I suppose, but do we really still expect such a prompt to have any effect, and can we measure it? It feels very 2023 “prompt engineering” and not the kind of thing you’d put any emphasis of in promoting a modern LLM. Stands out as surprising.
> The system prompt used in the standalone compaction request is different. Instead of telling the LLM "you are an expert coding assistant", we tell the LLM "you are a context summarization assistant."
No disrespect to Pi or the team, it’s my daily driver, I have a lot of respect for the project. But this feels naive to me. You might as well do it I suppose, but do we really still expect such a prompt to have any effect, and can we measure it? It feels very 2023 “prompt engineering” and not the kind of thing you’d put any emphasis of in promoting a modern LLM. Stands out as surprising.