@flavioludi Test it out for yourself. Build a small app or feature by hand, and then try again using an agent as the driver.

To me, it still feels like babysitting an LLM that will never really learn the fundamentals of what you have just asked it to fix.

It's more work (and more stressful) than just writing the code.

But yes, I know that in the AI era, if someone says that using AI is not that much productive, people will immediately say that you are doing something wrong.

I've tried Vibe coding and I have no clue what's going on. I think Vibe coding is difficult to scale but I don't have enough experience with it to really say Vibe coding is a dead end or not at this time. You need to keep AI on a really tight leash or it just kind of goes off in some random direction doing it's own thing.

How should we write docs for such architecture to help other developers in future? Uml sequential diagrams?)