It's surprising how low-level the conversations are about AI-assisted coding from people who've never really gone deep with it themselves. It reminds me a bit of those 1970s debates about whether kids should be allowed to use calculators for math.
Just because some folks haven't gained real experience with these new coding tools doesn't mean experienced developers with decades under their belt don't have it. I didn't have any at the beginning either. It was a painful process figuring out what worked and what didn't, and building up all my safeguards in a project. I've fallen into so many traps where I had to rebuild everything because the AI recommended the wrong tool or the wrong method due to its lack of experience.
I came close to wanting to give up completely 3 different times with #NoorNote because the AI had messed things up so badly for me. I even restarted after a month, luckily pretty early on.
But now I know exactly what works with AI and what doesn't. I know when I need to dive in deep myself and where I can give it more freedom. The whole system of safeguards and controls I've built around it is worth its own write-up, and to me that's what actual AI competence really looks like. Because only this way can one person actually ship a big project in 9 months with AI, even if it took a painful learning curve, something a 5-10 person team still isn't capable of pulling off in a year.
The next person who starts talking shit about AI needs to show some real proof-of-work first before I take them seriously.