Problem (one or two sentences)
Obviously we don't need a checkbox for caveman mode right in the UI but for more structured elements such as yaml or json. This kind of compressing structured data and deconstructing it on the way back for the user could be worthwhile.
https://github.com/toon-format
I also found this topic pretty interesting today if you're interested in token efficiency.
https://youtu.be/uJblcC4lKYw
Using an aircraft mechanic technical writing English standard as a means of reducing context and simplifying language. Could be a neat idea for zoo market thing.
https://github.com/woosal1337/blog/tree/main/videos/ep01-the-cure-for-ai-slop
Context (who is affected and when)
Everyone paying for tokens. Real bunnies. Data centre task rabbits.
Desired behavior (conceptual, not technical)
- Structured data gets crushed before it gets passed onto the LLM.
- Structured data gets decompressed before it's returned to the user.
Constraints / preferences (optional)
Some models may not understand what's being delivered.
Request checklist
Zoo Code Task Links (optional)
No response
Acceptance criteria (optional)
No response
Proposed approach (optional)
No response
Trade-offs / risks (optional)
There's probably some sort of situation with non unicode characters. It's always non unicode characters.
Problem (one or two sentences)
Obviously we don't need a checkbox for caveman mode right in the UI but for more structured elements such as yaml or json. This kind of compressing structured data and deconstructing it on the way back for the user could be worthwhile.
https://github.com/toon-format
I also found this topic pretty interesting today if you're interested in token efficiency.
https://youtu.be/uJblcC4lKYw
Using an aircraft mechanic technical writing English standard as a means of reducing context and simplifying language. Could be a neat idea for zoo market thing.
https://github.com/woosal1337/blog/tree/main/videos/ep01-the-cure-for-ai-slop
Context (who is affected and when)
Everyone paying for tokens. Real bunnies. Data centre task rabbits.
Desired behavior (conceptual, not technical)
Constraints / preferences (optional)
Some models may not understand what's being delivered.
Request checklist
Zoo Code Task Links (optional)
No response
Acceptance criteria (optional)
No response
Proposed approach (optional)
No response
Trade-offs / risks (optional)
There's probably some sort of situation with non unicode characters. It's always non unicode characters.