By boosting the accuracy and frequency of the reading you can get pretty good results.
But that has little to do with LLMs, or LLM generated code.