Identify the assumption that kills your idea if wrong, design 3-4 tests that produce real signal, and build a 2-week go/no-go playbook — before writing a line of code.
