[zachrose]

I second this. Additionally, if the choices are still overhelming and you have an idea of what you really want to build, I would suggest trying something like writing cucumber features[1] and running them in a headless browser.

I'm sure many people will tell you about their negative experiences and how writing these kinds of "tests" are slow or misleading or a waste of time. But I would say that deciding what your software does--even in an abstract and gradual fashion--gives you tremendous freedom in knowing that all the other choices are just implementation details of a larger design. (As the parent says, a "lower level of commitment".)

[1] https://cucumber.io