TABARC-Code

just a quick one really. Thought would share. I noticed how it treats language and its habit of using grammar in a traditional sense. 

The danger of English is we have a tendency to be very loose with it and because of this l, our use of the English language is also at times sloppy in how we ask questions or how we propose questions to an LLM so this was actually a bit of an idle thought that I tried and noticed significant improvements. 

I didn't use an LLM to restructure the question because it's the danger of the feedback loop 'if you ask a sloppy question or a badly worded question' initially even without meaning to, you create a negative feedback loop that the AI helpfully tries to correct by also using sloppy language but if you're careful in how you structure the question yourself then suddenly your initial query is correct and therefore you can then use that as a positive feedback loop in order to improve future outputs. 

so the use of words like Read this, or Audit this perform similar results but