When vibe coding sometimes i’ll be asking for composer to do something (i.e. create a new feature or review how a section of code works) but my idea will be slightly scattered meaning the prompt isn’t properly (ir optimally) structured for the model to properly execute what i’d like or the core of what my prompt is asking -
Having a ‘optimize prompt’ button which distills whatever is mentioned in the user’s prompt and structures it in an optimal format for an instruction model would
-
Result in more useful, effective outputs from the model leading to faster results when using Cursor
-
It would likely save Cursor inference costs from having users iteratively have to redefine EXACTLY what they want the model to do unnecessarily
The optimize prompt could even be based on benchmarks of the best way to structure information to models to get the most coherent outputs - (but that’s more something for the Cursor team to worry about).