Vibe coding · Debugging basics · For beginners

How to Read Error Messages When You Can’t Code

The red wall of text isn’t yelling at you. It’s pointing.

Three questions, six real errors, and the one line you can safely ignore.

You asked the AI for one new feature. You run the app. And there it is: forty lines of red text, half of it paths you have never seen, ending in a word like TypeError that sounds like a diagnosis.

So you do what almost everyone does. You select everything, paste it into the chat, and type “fix”. Sometimes that works. Sometimes the AI changes four files and the error turns into a different error, and now you have no idea what you are even looking at.

Good news — you do not need to understand the code to read the message. An error message is a short report with three parts, and you only need to find them. If the app was already working and you just want it back, start with how to go back with git first, then come back here.

The three questions that decode any error

Every error I triggered answered the same three questions. Ask them in this order:

1. What happened?

This is the error name and message, like ReferenceError: itmes is not defined. The name (ReferenceError, TypeError, SyntaxError) is the category. The message is the detail. MDN keeps a reference of JavaScript error messages if you want the official wording for any of them.

2. Where did it crash?

Look for the first line that points at a file you or the AI wrote, something like cart.js:2:16. That means file, line 2, column 16. Lines that mention node:internal or a folder called node_modules belong to Node and your installed packages, not to your code.

3. Who called it?

The lines around the crash show the chain of calls that led there. This is where the real mistake usually hides, and it is the part most beginners never read.

One more thing before the examples: the two languages print errors in opposite order.

Example 1: the crash that blamed the wrong line

Here is a tiny file in the style of something an AI might write. It prints a receipt for a user:

function getTotal(order) {
  return order.items.reduce((sum, i) => sum + i.price * i.qty, 0);
}
function showReceipt(user) {
  const total = getTotal(user.order);
  console.log(user.name + " owes " + total);
}
showReceipt({ name: "Mina" });

Running node cart.js printed this (I trimmed the Node-internal lines and shortened the paths):

cart.js:2
  return order.items.reduce((sum, i) => sum + i.price * i.qty, 0);
               ^

TypeError: Cannot read properties of undefined (reading 'items')
    at getTotal (cart.js:2:16)
    at showReceipt (cart.js:5:17)
    at Object.<anonymous> (cart.js:8:1)
    ... (7 node:internal lines)

Question 1: something called undefined was asked for its items. Question 2: line 2. But line 2 looks perfectly fine. Question 3 is the one that solves it: getTotal was called by showReceipt on line 5, which was called on line 8 with { name: "Mina" } and no order at all. The line in the error is where the program fell over, not necessarily where the mistake was made.

Which line is actually the problem? (syntax errors)

Syntax errors are the cases where the language cannot even read the file, and they are the most misleading of all. Two real ones:

$ python3 loop.py
  File "loop.py", line 1
    while True print("Hello")
               ^^^^^
SyntaxError: invalid syntax

Python points at print. The mistake is the missing colon just before it. The Python docs describe exactly this: the error is reported where it was detected, and the real fix can be right before that spot (see the Python tutorial on errors).

$ node broken.js
broken.js:5


SyntaxError: Unexpected end of input

My file had a function with a missing closing } on line 3, but Node reported line 5, the end of the file, and showed nothing under it. For syntax errors, read the line the error names and the lines just above it. A missing bracket is an easy slip whenever code is edited in the middle of a file.

The error that is not a code bug — and the easy fix

Error: Cannot find module 'express'
Require stack:
- app.js
  ... {
  code: 'MODULE_NOT_FOUND', ... }

This one looks scary — and it is the friendliest of the six. The code was fine; the package it imports was simply not installed on that machine. I ran npm install express in the same folder, ran node app.js again, and it printed started. If an AI hands you code that imports something new, this is an error you may well meet.

A Python traceback, read from the bottom up

Traceback (most recent call last):
  File "app.py", line 7, in <module>
    print(checkout([{"price": 5}, {"cost": 7}]))
  File "app.py", line 5, in checkout
    return sum(load_price(item) for item in cart)
  File "app.py", line 2, in load_price
    return data["price"]
KeyError: 'price'  

Bottom first: KeyError: 'price' means a dictionary had no entry called price. The frame right above says it happened on line 2, in load_price. The frames above that are the callers. Line 7 shows the data that went in, and there it is: the second item says cost, not price.

Here is where vibe coding adds a trap. A quick AI fix might silence the error by making the code accept missing prices. That would make the red text disappear and hide the bad data. Ask the AI why the value is missing, not only how to stop the error. That is my advice from how this played out in a small test; I have not measured it across AI tools.

What to paste to the AI (and what to check first)

Pasting everything is not wrong, but pasting the right things gets you better answers. A short checklist:

• The command you ran (for example node cart.js), so the AI knows where the error came from.
• The whole error: for Node, from the file:line header down through the first few at lines; for Python, the full traceback.
• The file named in the error, and any file the call chain passed through.
• What you changed right before it broke. This one fact saves the most time.

Then add one sentence: “Explain what this error means in plain English before you change anything.” That is my suggestion, and I have not tested whether it improves results in any particular tool. It does make you read the explanation, which is the point of this post.

The honest limitations

I only tested Node and Python. Errors in the browser appear in the console (right-click the page, choose Inspect, then Console). Their text is similar to Node’s, but I did not test that here, so use the MDN error reference for the exact wording.

Real errors can be much longer. My test files were tiny, and I did not test bigger frameworks. The three questions should still apply — but finding your own file in a long trace takes patience.

My “plain meaning” column is a simplification. It is how I would explain each error to a friend — not an official definition. When it matters, read the official page.

Reading is not fixing. Knowing what, where and who called it gets you a clear sentence to give the AI. It does not guarantee the AI will fix it on the first try. If it keeps going in circles, write the plan down before you prompt or give it standing rules in a CLAUDE.md.

Who should actually care

Quick answers

Do I need to understand the code to read an error?

No. You need to find three things: the error name and message, the first line that points at your own file, and the lines around it. You can do all three without reading the code itself.

Which end of the error do I read first?

In Python, the bottom: the last line says what happened. In Node, the error name sits near the top and the call chain runs downward from the crash. That is from my runs, and the Python rule is in the official docs.

Why does the error point at a line that looks fine?

Because the error shows where the program stopped, not where the mistake was made. In my test, the real cause was a missing argument three lines away, or a missing colon just before the flagged spot.

Should I paste the whole error into the AI?

Usually yes, together with the command you ran and what you changed last. Check it for anything that looks like a key or password first.

Written Nov 11, 2026. Updates will be logged here: none yet.

Related free AI guides

• AI broke my working code: how to go back with git
• Vibe coding planning guide: plan before you prompt
• Env file security guide: keep API keys out of your code