Curious if anyone is finding success using certain prompts to get something done for in-flight projects, for me personally I found success in having Claude, Cursor etc create architecture, flow-charts, and logic markdown files for me to reference later on if I'm digging into a specific error, or when I need to update a delicate feature. It's not bullet proof, but I also tend to use "As a senior Developer, Perform a Root Cause Analysis on the following errors xxxxx, and symptoms I'm noticing xxxxx do not act on it, provide contributing factors, rank these factors by likelihood, and create a remediation plan". The more I push the AI system to understand a problem holistically (if I can at least) the bugs, and error conditions are sometimes easier for me to isolate and have AI apply fix without breaking some logic elsewhere.
Again, none of this is guaranteed, just sharing my personal experience wrestling with in-flight builds that I'm expanding or adding on to and encountering bugs with. Curious, how do you all handle this? got any tips to share?