Reading Errors Like a Detective
The lecture, after the doing: how to read an error without knowing syntax, and why AI explanations still need checking.
🔎 You Don’t Read Errors — You Report Them
Last lesson you triggered two errors and fixed both, without translating a single line of technical text yourself. That’s not a shortcut you’re only allowed to take as a beginner — it’s the actual correct process, at any skill level, when you’re directing rather than writing code by hand.
A useful bug report has three parts, and you already produced all three last lesson without being told the formula:
| Part | What it means | What you actually said |
|---|---|---|
| What you did | The exact action that triggered it | ”I clicked the fun-fact button” |
| What happened | What you actually observed | ”Nothing changed” or “I got this error: [text]“ |
| What you expected | The outcome you wanted instead | ”It should show a new fact each time” |
That’s the whole detective method: not decoding the error’s internals, but reporting its shape precisely enough that someone else — in this case, the AI — can diagnose it.
🚨 Crash vs. Silent Bug
Not every problem announces itself. There are two different failure shapes worth telling apart:
- A crash — something visibly breaks: an error message, a blank screen, a feature that stops responding entirely. Loud, but easy to notice.
- A silent bug — the app still runs and looks fine, but does the wrong thing. The fun-fact button that always shows the same fact instead of a random one is a silent bug — nothing crashed, it’s just not doing what you asked.
Silent bugs are actually the trickier ones, because nothing forces you to notice them. This is exactly why the Test step from Module 2 matters so much: it’s your only real defense against a bug that never announces itself.
⚠️ Confidently Wrong Is a Real Pattern
Here’s the part of this lecture that matters most: AI models can explain a bug fix with total confidence and still be wrong, or only partly right. This isn’t a rare glitch — it’s a known pattern worth actively watching for, not something to trust blindly just because the explanation sounds fluent and technical.
If an explanation uses technical language you don’t recognize, it’s tempting to assume that means it’s correct — surely something that sounds that precise must be right. It doesn’t work that way. Fluency and correctness are two different things.
Your actual check is always the same one from Module 2’s loop: run it, and see if the specific thing that was broken is now fixed. Not “does this explanation sound convincing,” but “does the button now do what I asked.” That’s a test you’re fully qualified to run, regardless of how technical the explanation sounded.
The AI gives you a long, technical-sounding explanation for why your app broke, and says it's fixed. What should you do before trusting that it's actually resolved?
Reveal Answer & Explanation
Run the app and specifically re-check the thing that was broken — regardless of how confident or technical the explanation sounded.
A fluent explanation is not the same as a verified fix. The only real confirmation is testing the specific behavior again, the same way you did throughout Module 2’s loop.
⏭️ Up Next
Check how your relationship with error messages has shifted since this module started.