Sure, and that's why I explicitly said "experiment had a positive result". Before you decide a particular chunk of code is a keeper, you should have data that g

Sure, and that's why I explicitly said "experiment had a positive result". Before you decide a particular chunk of code is a keeper, you should have data that gives you confidence. Before you have signs of product-market fit, I think people should write throwaway code and then throw it away. Once you have something worth investing in, I think it's time to invest. Having high-quality code in areas where you're confident it's valuable enables rapid experimentation in the areas where you're not sure yet.