Chapter 5 · Part 2

Coding and getting things done

You don't have to be a programmer to get huge value here — ChatGPT writes spreadsheet formulas, explains error messages, automates repetitive chores and drafts plans. The rule that makes all of it work is the same: give it the specifics. Vague in, vague out.

Debugging: paste the actual error

The single most useful coding habit is pasting the exact error message and the code that caused it. "My code doesn't work" is unanswerable; the real error is often instantly diagnosable:

Give it everything it needs
This Python function throws the error below. What's wrong and how do I fix it?
I'm a beginner, so explain the fix.

Code:
[paste the function]

Error:
[paste the full error message, including the traceback]

Understand code you didn't write

Inherited a confusing script or formula? Ask for a plain-English walkthrough:

Explain, line by line
Explain what this does, step by step, like I'm new to it.
Then point out anything risky or that could be simplified.

[paste the code or spreadsheet formula]

Everyday tasks, not just code

The same skill applies far beyond programming:

💡Things people forget it can do
  • Spreadsheets — "Write an Excel formula that flags rows where the date is older than 30 days." Describe your columns and what you want.
  • Automate chores — a script to rename files, a template for weekly reports.
  • Plan and break down — "Turn this goal into a week-by-week plan with milestones."
  • Convert formats — messy notes into a clean table, a list into a checklist.

Work in small steps

Don't ask for a giant finished program in one shot — you can't tell what's wrong when it breaks. Build in pieces: get one function working, test it, then ask for the next. Feed results back: "that returned an empty list — here's what I ran." Iterating with real feedback beats one huge request every time.

⚠️Run it, then trust it

AI-written code can look perfect and still be subtly wrong — a mishandled edge case, an out-of-date library call, a plausible-but-fake function name. Always run it and check the result before relying on it, exactly like the "verify" rule for facts. If it touches important files or data, test on a copy first.

You can now get real work done with it. The last chapter makes ChatGPT fit you — and covers the habits that keep you out of trouble.