I hate to turn everything into a conversation about AI, but this essay maybe explains best what LLMs have done for me recently. Particularly the first part. I want to add a new feature, but I want to keep things clean. It needs tests, CI, documentation. It makes exploring new ideas a bit cumbersome, because code tends to create minor distractions that eat up time. You miss a semicolon, or you forget the order of the…
Are you giving the LLM your code and letting it add a new feature to see how it works? Or what? How successful do you find your method? Would you recode what you introduced using the LLM?
It works pretty well with Claude Code. Much better than cursor, which is a step up from copilot. Even with the same models, and I'm not sure why. I haven't really tweaked much around the AI tools, since I don't really know much about how they work.
I've just found that Claude Code somehow... just works. Out of the box, no MCPs, no fancy configs. I just straight up tell it what I want, and most of the time it gets it right or close enough to right that a second instruction is all I need.
Would I recode what I wrote? Maybe not from the ground up, since I already had a pretty good framework. But LLM has managed to make some pretty fiddly changes to my codebase recently. It would have taken me a long time, mostly in tedious edits.