Reading Code (Without Knowing It)

You Don't Need to Read Every Line

You don't need to understand every line of code the AI writes. But you need to skim it well enough to:

1. Verify it matches your requirements — did it actually build what you asked for?
2. Spot obvious problems — hardcoded secrets, missing error handling, weird patterns
3. Ask informed follow-up questions — "Why did you use approach X instead of Y?"

Think of it like being a film director watching dailies. You don't need to know how to operate the camera, but you need to recognize when a shot doesn't match the vision.

What to Look At

1. Function and Variable Names

Good code reads almost like English. Before trying to understand the logic, just scan the names:

function calculateShippingCost(weight, destination) {
  const baseRate = 5.99;
  const perPoundRate = 0.50;
  const internationalSurcharge = 15.00;
  
  let cost = baseRate + (weight * perPoundRate);
  
  if (destination === "international") {
    cost += internationalSurcharge;
  }
  
  return cost;
}

Even without knowing JavaScript, you can read the names and understand: it calculates shipping cost based on weight, with a surcharge for international orders.

If variables are named x, temp, or data — that's a yellow flag. Ask for more descriptive names.

2. Structure, Not Syntax

Don't worry about semicolons and brackets. Focus on:

• How many functions are there? (More = better organized)
• Are there comments? (AI should explain its code)
• Is there error handling? (Look for try, catch, if error)
• How long is each function? (50+ lines = probably doing too much)

3. Hardcoded Secrets

Any time you see something like this, stop immediately:

# 🚩 RED FLAG — hardcoded credentials
database_password = "mySecret123"
api_key = "sk-abc123xyz789"

These should never be directly in the code. Tell the AI:

You hardcoded the API key directly in the code. 
Move all secrets to environment variables and use a .env file.

4. TODO Placeholders

AI sometimes leaves placeholder notes instead of finishing the work:

# TODO: implement proper authentication
# TODO: add error handling
# FIXME: this is a temporary workaround

If you see these, the AI handed you unfinished work. Ask it to complete everything.

Reading Error Messages — They're in English

Error messages look terrifying but usually tell you exactly what's wrong. Read them from the bottom up:

Traceback (most recent call last):
  File "app.py", line 42, in process_user
    user_data = get_user(user_id)
  File "database.py", line 15, in get_user
    return db.query("SELECT * FROM users WHERE id = ?", user_id)
TypeError: query() takes 1 positional argument but 2 were given

Common Errors in Plain English

Real-World Scenario: The Silent Failure

You asked the AI to build a registration system. The code looks right. When you test it, you see "Account Created!" But no user appears in the database.

You investigate and find this:

app.post('/api/register', (req, res) => {
  const { name, email, password } = req.body;
  
  // TODO: save to database
  
  res.json({ success: true, message: "Account Created!" });
});

The AI left a TODO and never wrote the database logic. It returns a success message without doing anything. This is a real and common failure mode — code that looks complete, runs without errors, sends success responses, but the core functionality was never implemented. (This is the same silent failure pattern from the walkthroughs.)

The Big 3 AI Security Risks

These three issues appear in AI-generated code more than anything else. Make them your default checklist every time you review output:

Quick pre-install check:

Before I install these dependencies, verify that each package name
is correct and exists on [npm/PyPI]. List the official package URL
for each one.

How Much Should You Understand?

When You're Completely Lost

If you're staring at code and none of it makes sense — don't panic. Paste this:

Explain this code to me like I don't know how to code.
For each section, tell me: what it does, why it's there,
and whether anything looks risky or incomplete.

The AI wrote it — it can explain it. If the explanation doesn't make sense either, that's a sign the code might have problems. Bring in a second opinion.