ever since around December it went from 20% to basically 95%+ the advancements in agentic loops + models have made it kind of silly not to use it for everything. I still prefer to be “in the loop” and ask the ai to change around 30% of edits, so i’m still writing the code in some sense, but its more like pair programming than vibe coding for me. I still think vibe (hands off) coding produces bad results. also for me its only more productive if i have multiple agentic loops running in parallel, hence why i built Damus Agentium to handle hopping between edits automatically

Depends a ton on what for. For small applications, scripts, tests, self-contained things, or repetitive tasks (think refactoring), it’s basically 100% AI+review (often tests you have to tweak and carefully check cause it’ll write tests that pass but don’t actually test anything with it being obvious at all!). For high-level tasks in large codebases, especially starting to design a new API or feature, it’s somewhere between 0 (doing it yourself to explore an approach while being more directly exposed to the changes) or using them to explore but then rewriting chunks after you’ve explored different options.

I don't consider myself a "professional" (serious) coder these days, but to answer your question: it changed gradually and then very quickly. Over the last 6 months I'm pretty sure I wrote zero lines without prompts.

I use AI very heavily while I program, but I'm still the one programming. I think "vibing" is cool for non-programmers. I think professional programers who are going autopilot (or writers or lawyers or...) are playing with fire. It can turn you into a slop zombie or make you the best you've ever been, but that's up to how you use it.

I tell it what needs to happen and define the success criteria, and poke it when it gives up Ive written 0 lines of code and won't type at all, even for one line changes like constants