Re:Show HN: Cursor AI Rules Directory (Open Source)
by Kiro on Sunday August 25, 2024 @09:29AM
(#41347158)
Are these system prompts actually adding any value? I use Claude for coding a lot and haven't seen any meaningful difference from just giving it the code and asking for a change. Telling it to "be an expert" in something feels like a type of thing that sounds cool in theory but has no actual impact in practice.
[ Parent
| on HN ]
Re:Show HN: Cursor AI Rules Directory (Open Source)
by mamp on Sunday August 25, 2024 @09:46AM
(#41347279)
I suspect it's not so much the "expert" instruction but the list of subjects of the expertise. These words will generate embeddings that have a better chance of activating useful pathways and relationships within the LLM's generation path.
My understanding is the goal is to prime the LLM with context that it will use when generating answers, rather than hoping it will infer connections in the feed forward layers when generating answers based on a sparse prompt.
Re:Show HN: Cursor AI Rules Directory (Open Source)
by pontusabra on Sunday August 25, 2024 @09:32AM
(#41347181)
it does, especially when you have libraries that you use and want to instruct how to write the code!