[pbecotte]

If the accepted wisdom was true, that you can trade quality for speed, then you shouldn't write tests.

I do not believe that to be the case. I find the biggest value of tests is being able to change the code more quickly. Running tests is MUCH faster than a regression script.

There is a scale where you'll never have to change the code enough for the tests to pay off. Personally, I think that scale is quite small, maybe a week of development work.

However, it takes some experience to get good enough at the tests that you feel that benefit. If you're not there, the answer is probably different. However, if your startup scale is non trivial, I'd bet that the payoff if learning that skill and building some testing framework from the get go would be worth it.