We need fewer assumptions to test

with each deployment (smaller changes), not more delayed feedback (more gates, more branches, etc.).

It’s hard to stick to testing fewer assumption if getting to test one takes disproportionately long time, but that’s not the reason to start testing more assumptions at once, which aggravates the problem even more.

It’s a reason to reduce the cost of testing assumptions by getting faster, not slower, to production in smaller steps.