Compress your prompt below. Optional if your prompt is short and direct. For prompts over 1500 chars, use Haiku with the compress.txt system prompt from GitHub.

Paste the compressed prompt into your expensive model (Claude, GPT, Gemini Pro). Include the think.txt system prompt so the model compresses its output. Get it on GitHub.

Copy the compressed response and decompress it below.For longer outputs (over 2000 chars), use Haiku with the decompress.txt system prompt / project instructions from GitHub. FOR CODE: Add the CODE_LEAN EXTENSION extensions/code_lean.ext to the project / system prompt

Continue the conversation: paste your follow-up into Step 1 (or type directly if short), compress, send to expensive model, decompress. Rinse and repeat.