gist.
Fewer words.
Same meaning.
Remove the excess.
Keep the thought.
Gist would use an LLM to rewrite ordinary English as compact, understandable English. It is not a new vocabulary, and not a shorthand that trades away precision for speed.
One idea. Two forms.
I would like you to create a website where users can sign in using Google accounts. Each user must only be allowed to edit their own notes, and those notes should be private by default.
Build website. Auth: Google. Users edit only own notes. Notes private default.
Same ownership and privacy requirements, expressed more directly.
Illustrative rewrite, not the output of a tested compressor. Actual token reduction and comprehension still need measurement.
Meaning first.
Less language
Remove redundancy, filler, and unnecessary grammatical scaffolding.
Same intent
Retain facts, conditions, negation, names, numbers, and requirements.
Measure it
Test understanding, token counts, and the total cost of compression.
The first prompt.
Start with an ordinary model and a short, precise instruction. Then measure what survives.
Compress English for LLM comprehension.
Goal: minimum tokens, maximum meaning.
Rules:
- Use ordinary English words.
- Remove filler and redundant grammar.
- Preserve facts, names, numbers, negations.
- Preserve conditions, exceptions, obligations.
- Never invent or omit requirements.
- Output compressed text only.A compressor is only useful when downstream models understand the message and the saved tokens outweigh the cost of compression.