←back to thread

577 points Delgan | 2 comments | | HN request time: 0.475s | source
1. rpicard ◴[] No.44352003[source]
This would be a cool place for LLMs to store a summary of the prompts used to generate the code in order to make it easier for other LLMs and humans to pick up where they left off.
replies(1): >>44352928 #
2. fudged71 ◴[] No.44352928[source]
I've been thinking about this exact same problem. It would be great to store a log of how much time/tokens were spent reasoning, what was the reasoning path, why were certain decisions made, etc.

I don't know if rationale is something better suited for the git commit log, or tagged by code function to an external "rationale" system of record.