Comment by logicchains

It's hard to see where you're coming from. With o1 Pro or Gemini 2.5 Pro I can give a detailed text specification of a module I want to write to it and it'll write code to do exactly that, with fewer errors than I'd make on a first try. It'll also generate unit tests for it. Writing a text specification then reviewing some code for me is way, way faster than writing all that same code by scratch.

Are you working on code that's heavily coupled to other code, such that you rarely get to write independent logic components from scratch, and IP restrictions prevent you from handing large chunks of existing code to the LLM?