Debugging and Performance
From mystery bug to resolved with regression tests
9 lessonsPlus
Lesson 1 is free now. The rest comes with Plus.
What's inside
- AI-assisted root cause analysis
- Writing regression tests
- Documenting bugs for team learning
What you'll gain
Track down and fix production issues faster using AI as your debugging partner.
- You'll systematically identify the root cause, fix the issue, write regression tests, and document what happened so the same class of bug doesn't re-appear.
When to use it
- A bug is reported in production
- Performance degradation is detected
- Need to prevent similar issues from recurring
You'll reach for it: Every bug or performance issue
What you leave with
6
Prompts
Syllabus
9 lessons1
From Bug Report to Failing Test
You can't fix a bug you can't trigger. We'll get the agent to turn a vague user report into a failing test that reproduces the bug every time, so you'll know for certain when it's actually fixed.
3 minFree
2Fix the Cause, Not the Symptom
It's surprisingly easy to make a failing test pass without actually fixing the bug. We make the agent explain the root cause, with the exact lines, before it touches any code, so you get a proper fix and a free regression test, not a patch that hides the problem.
3 min
3Never Fix the Same Bug Twice
Fixing the bug is only half the job. Once the fix is in, we work out what would have caught it and adjust our environment so the same kind of bug never makes it back in.
7 min
4Corner the Bug You Can't Reproduce
Some bugs trigger perfectly every time you try to reproduce them, and some don’t. We'll corner one using an instrumentation loop, where the agent adds a few targeted log lines, reads them back, and narrows things down one hypothesis at a time until we've found the trigger.
4 min
5Let AI Find What's Slow
Guess at a performance problem and you'll usually optimise the wrong thing. We point the agent at the problem instead, and it hands back a ranked report of what's slow, why, and how to fix it, all before we touch any code.
2 min
6Fix It, and Prove It's Faster
A performance fix without a measurement is just a guess. We work through our report fix by fix, proving every one with the same request timed before and after.
7
Catch Slow Code as It's Written
Fixing slow queries doesn't stay done, there's always another N+1 on the way. We switch on strict mode so slow code gets caught the moment it's written.
3 min
8Know When Production Breaks
Most production errors never get reported. We'll wire up Sentry so errors report themselves, with the stack trace, the request, and exactly what the user did beforehand.
3 min
9From Alert to Shipped Fix
An issue reported using a bug reporting service is the best report you'll ever get, and the agent can read every bit of it. We take a real production error from the alert landing all the way to a shipped fix, then resolve it against the release so we know it actually worked.
4 min
Access
It's part of Plus. Lesson 1 is free to watch for everyone, no account needed, and so is the brief above.
Every lesson of every course, the full notes, transcripts and prompts, and the Unlearn-built tools. Blog posts and the tools that are public on GitHub stay free for everyone.
It shows as Completed on your profile and counts toward its role. Finishing every step of a role earns the role certificate.
Yes. Team plans can assign this course, or a whole role, and follow progress in Team.
Next step
Next in Product Engineer
Refactoring
Workflow · 9 lessons · 56 min