Hi HN - so my current dev workflow using AI assistance is basically: Give a spec to chatgpt, copy generated code into my IDE, run code, inspect output, then iterate by asking chatgpt for additions, changes etc, re-pasting into my IDE. The problem with this is that I often want to make my own changes to code directly in the IDE, and as such what's in my IDE gets out of sync with what's in chatgpt. Am sure this is a solved problem, so I was wondering what people's preferred approaches were for integrating LLMs into their dev workflow?
Comments
sgt•3mo ago
I think a lot of people use Cursor for this reason.
I use JetBrains IDE's and AI Chat, which is able to look at your files, but it doesn't change your files. It'll propose new changes or full source code files when needed. A bit safer and it's fairly efficient to me.
When things change in the IDE you just start a new chat to avoid too much context that is now out of date.
sgt•3mo ago
I use JetBrains IDE's and AI Chat, which is able to look at your files, but it doesn't change your files. It'll propose new changes or full source code files when needed. A bit safer and it's fairly efficient to me.
When things change in the IDE you just start a new chat to avoid too much context that is now out of date.