@metaend.eth
Compressed my AGENTS/md by 41% with zero information loss.
Same file. Same instructions. Same behavior from the AI agent. Just 59% of the original tokens.
Here is the exact prompt I used:
# Token Efficiency
- No articles (a, an, the)
- No filler words (just, really, very, basically)
- No pleasantries or preamble
- Shortest possible names in explanations
- Use symbols over words (-> not "returns", => not "leads to")
- Code over explanation when code is clearer
- One sentence where others use a paragraph
That is it. Copy, paste, done.