Super cool! Tried it on my prompt which tried compressing information using emojis, but those were given a low importance score. Switched the emojis out for plain text, which is given a higher importance score, and I'm seeing better results.
HN user
naguas
3 karma
Posts0
Comments4
No posts found.
Show HN: A surprisingly effective way to predict token importance in LLM prompts 3 years ago
Currently waiting on 3 PRs to get reviewed just so I can start developing on top them. This is a major pain point in my current development. My workaround is to have two separate copies of our monorepo locally and develop on both of them since our internal git wrapper doesn't allow two PRs from the same file. It probably wouldn't be that hard to extend our git wrapper to call graphite. Just signed up for the waitlist!
Very well done website! I love the detail on the tech and audit.
Anki: Memorization with Spaced Learning 8 years ago
I started Anki about two weeks ago and I've already reviewed over 1,200 cards. It's a fun way to start the day and is surprisingly therapeutic. It really makes remembering a choice.