@gabrieljkeller Here’s a less caveman-flavored token reduction experiment. https://t.co/Rq1iF3U7Rs It still affects output quality depending on the model.
Claude Code’s leak includes KAIROS, their unreleased always-on memory-consolidating daemon. I built something similar during my undergrad research. It’s live today, open source, and works with any AI coding setup.
https://t.co/SSqw1XcB2h
I built a chatbot at #TreeHacks this weekend that predicts whether you’re sick, stressed, or intoxicated solely based on the sounds of your voice and words https://t.co/jbMR5CyDN8