Debugging methodology proposed by Maciej Rząsa at wroclove.rb 2024, framed as borrowing the scientist's workflow rather than their results. Loop: gather observations and read relevant docs/source ('library' work), formulate an explicit hypothesis with supporting evidence, design the simplest possible experiment that could disprove it, run the experiment, update understanding, and publish findings (notes, Slack, postmortem). A bug is considered fixed when you understand what happened — not merely when the symptom disappears. Scales to a team when every member publishes small observations and hypotheses so others (even in other time zones) can pick up the trail. Named 'hypothesis-driven development' as a tongue-in-cheek play on the '<X>-driven development' genre.