Raw LLM Responses

Agreed, 90% of my code is not written by hand. The code is just as good as I would have written. I know a lot of people will use this to hurr durr says stuff about your code quality, but yeah, no it's really good, it's just the reality. The good news is software engineers value isn't just in mechanically outputting code, but rather solving problems, and this as a tool to help develop complex systems has been working great for me. I've also said it a lot, but Rust seems to be one of the reasons I am having such a good experience. The compiler stops the hallucinations. The compiler diagnostics, clippy lints, etc. all feed back to the AI to generate code. We already have a huge corpus of pre-AI written Rust internally so following similar style that my team has had for years is not too hard for it to figure out. Other things I use it for: - General search - Drafting technical documents - Generating excalidraw documents for my design docs - Reading an unfamiliar codebase and helping me trace features - Good ole println! debugging, something tricky happening in prod, can you insert a bunch of trace prints to figure this out - Unfamiliar service debugging, here's a stacktrace, any thoughts? - code review - Figure out what log query I have to do to get some information from a service Every single thing there I can do without AI, and have plenty of times over my career. The AI is just something that can do it really quick. And sure sometimes it does something weird, I just redo and tell it why that wasn't what I expected.