Show HN: Keep large tool output out of LLM context: 3x accuracy 95% fewer tokens (github.com) 7 pts| 1 day ago | 1 comment
