I TDD all my code...when the project is stable and stablished.

TDD is usually a waste when you are exploring, and don't even know what you want to do. Let alone TDD won't get you any users, so you should spent the time on that first. Otherwise you'll end up with nice tech no one uses, which is a nice hobby but a terrible business.

You should also be ready to throw the prototype or MVP through the window and rebuild it if you get funding anyway, or reach a point where your original simple architecture doesn't accommodate for the new scale AFTER you get enough actual users. So unless TDD is helping you for real while developing certain things (mostly algorithmical), I'd skip it.

Once the project is mature, sure, TDD all the way, I'm a 100% proponent for this.