Thanks for an interesting read! I can relate to this.<p>Putting effort into tools to increase diagnostics comprehension pays off in more quickly solving bugs. However, prior to AI, building specialized tools (or even trying to find and learn existing ones) introduced a threshold.<p>I did exactly what you stated:<p>> “Without AI support, I would have just muddled through with my existing tools, lacking the time or focus to build a debug UI.“<p>I once took a greater initiative to make my debugging more manageable instead of struggling with the limitations of existing default tools.<p>In my case PHP/DB/AJAX interaction opened for hard to trace bugs; the initial challenge was to first target the right level for diagnostics. This was pre-AI, and I wrote my own diagnostics tool. It worked. Unfortunately the need wasn't in proportion to the effort I spent building the tool as my project ceased to be relevant.<p>Now my environment is Node JS and database (backend), and JS (front-end) as well as the server infra. I had AI create node js server accessible by websocket, unifying diagnostics from all layers. Then a simple web UI with coloring and ability to filter by source etc. Then layer-specific tools to allow layers to send diagnostics. Even my local Nvim diagnostics are sent there.<p>This was a breeze to put together and allows me to remain focus on my main goal. A bonus part is I actually do learn along the way, as the iterative AI process sparks curiosity about the system. Often I find myself in joyous exploration with the AI.