I really disagree with this. There's a point on most projects where you've learned about all you're going to learn from it. Maybe it's still growing, but it might not be teaching you anything new. The best way to get good is diversity of experience, and the best way to get diversity of experience is to do a lot of different things. The more diversity you can pack into a smaller time space (without just cramming and rushing things), the faster you'll get good.
If you've only been programming 3 years, you're definitely going to make a lot of mistakes. That's good! That means you're pushing your boundaries and learning. Peer review feedback isn't a mark of shame; and being attached to your code is a _bad_ thing. At the end of the day, coders vastly overvalue code beauty and aesthetics, and overestimate how long their code is really going to live. (I've been a professional for about 11 years, and while I take pride in my work, a lot of the companies I worked for either folded, pivoted to a new product, replaced some of the things I wrote with open source solutions after the problem space had become less novel and a proprietary solution didn't make sense anymore, or a billion other reasons why the code didn't need to live anymore. I'm not saying that's an excuse to write shabby code, but a lot of times "it works and it solves the problem at hand and the code isn't a disaster" is when you should stop working on it).
I think coding is a lot more like creative writing than it is like engineering, and if you want to get good at creative writing you make a point of writing a lot of stories, even though a lot of them won't be good. Or if you're composing music, you write a lot of music, you don't focus on one piece forever. Anything where you make, the more you make, the quicker you get better at it.