That's cool! One thing I would like is to have an `llm` agent that takes your current un-staged files and automatically figures out how to put them into different commits.
This technique might be more efficient but can be highly correlated to the order of the input text. The paper [1] I mention in the repo touches upon such methods briefly.
It's astoundingly less efficient right? How many compares ( and LLM calls ) to rank 10 items in order? And is it actually stable? You could get a ranking with logprobs in one llm call for 10 items, or do it n=3 times, with a shuffled order and average them out. I'm not sure how to scale to larger sizes of items though.
I guess it depends on how many items you are sorting, but when I think about sorting I think about putting 100+ items in order.
Yes, that's correct! All three techniques have their trade-offs. I like the term "stability" you are using for this. There's definitely interesting avenues to go about this trading number of comparisons vs. stability.
