We're sorry, but something went wrong
summary
Advocates using the scientific method for debugging: start from a known state (branch or git stash), reproduce the failure with a failing test, form a simple hypothesis without conjunctions, and run one five-minute change at a time. Use git bisect, guard/autotest, rubber‑ducking or teammate pairing, and set timeboxes (30 minutes for beginners). Triage, stabilize (rollback, feature switch, or manual DB fix), submit a compact bug with minimal reproduction steps and a failing test, then run a factual post‑mortem with five whys.
AI-written summary · about
chapters (36)
- 0:00Start
- 3:22Identify A Problem
- 4:29Think Layers
- 4:45Make A Hypothesis
- 5:30Avoid Conjunctions
- 5:55Test Your Hypothesis
- 6:09One Thing At A Time
- 7:05Is It Fixed?
- 7:24a: Refactor
- 8:14Revert Changes
- 8:40b: Try Again
- 8:54Refine The Problem
- 9:25Tools & Methods
- 12:03Ask For Help
- 12:39Documentation
- 13:12Issues & Bugs
- 13:54Mailing Lists
- 14:17Stack Overflow
- 17:03Submit A Bug
- 19:24Minimal Repro Steps
- 19:42Related Cases
- 20:01A Failing Test
- 20:47Provide A Patch
- 21:21Feature Request
- 23:13Recognize The Real Problem
- 24:33Multiple Problems?
- 26:13Manual Hackery
- 26:58Resources
- 27:31Write A Test
- 27:42Scientific Debugging
- 28:18Expectations
- 28:52Improving Estimates
- 29:18Communicate Often
- 30:14Re-triage
- 30:305 Whys
- 31:07Debug With Science
related talks
26mThe Mindset of Debugging
26mA Practical Taxonomy of Bugs
26mDebugging Clojure Services Under Stress
14mYour Agent Failed in Prod. Good Luck Reproducing It.
42mBeyond “Just Try Harder”- Three effective strategies to tackle bugs
41mDetective Hat: Investigating Production Issues
21mLessons From Long-Running Investigations
31mPost-mortems: Building better software together.