I have been using cursor for about 2 months and grateful for the significant productivity gains. But from time to time I get into coding quagmires and the models seem to run around like headless chickens. I create instructions, refine prompts and start new sessions (resets) yet still too many iterations and frustration (clearly my thresholds changed). I noticed that if I copy the context (as per my understanding) to ChatGPT of Claude using the same model (?) these are able to blaze through the issue providing insights and understanding coding patterns better, always resolving the issue. I am not sure why that is the case. What do you think? Additionally, I have resorted to creating clear documents of resolutions of problematic patterns experienced in Cursor with ChatGPT and saving them in my project so that the model can reference them to avoid repetition of frustrating iterations.