When not to trust AI, and why fundamentals still matter

AI tools can write a working React page in minutes. That makes it tempting to think fundamentals no longer matter. The opposite is true. The faster code appears, the more it matters that someone can judge it. That someone is you.

This lesson explains where AI is weak, how to spot when it is wrong, and which skills to keep building so that AI makes you faster instead of making you dependent.

A language model produces text that is likely to follow your prompt. It does not run your code, and it does not 'know' things the way a textbook does. Most of the time likely text is also correct, which is why it is so useful. But sometimes likely and correct are different. Then you get a confident wrong answer, with no warning in the tone.

For how this works in more detail, see *How large language models work* in the AI Foundations track.

Here is a short function of the kind AI writes often. It looks correct and may even seem to work. See if you can find the problem before reading on.

// src/cart.js
export function addToCart(cart, itemId) {
  const existing = cart.find((line) => line.item_id === itemId);
  if (existing) {
    existing.qty += 1;      // changes the old cart object
    return cart;            // returns the SAME array
  }
  return [...cart, { item_id: itemId, qty: 1 }];
}

The first branch changes an object inside the old cart and returns the same array.

// inside a component
const [cart, setCart] = useState([]);

function handleAdd(id) {
  setCart(addToCart(cart, id));
}

React decides whether to re-render by checking if state is a new value. When addToCart returns the same array object, setCart sees no change, so the screen may not update when you add an item that is already in the cart.

This is the rule of immutability in React state: never change state in place; create a new array or object. A person who knows this spots the bug in seconds. A person who does not will paste the symptom ('the quantity doesn't update') to the AI, get another guess, and possibly lose an hour. The fix:

// src/cart.js (fixed: returns new objects, never edits the old cart)
export function addToCart(cart, itemId) {
  const existing = cart.find((line) => line.item_id === itemId);
  if (existing) {
    return cart.map((line) =>
      line.item_id === itemId ? { ...line, qty: line.qty + 1 } : line
    );
  }
  return [...cart, { item_id: itemId, qty: 1 }];
}

map builds a new array, and { ...line, qty: line.qty + 1 } builds a new object for the changed line. Unchanged lines are reused.

The point is not that AI is bad. It is that the fastest way to be correct is to understand the language, the framework and the data flow well enough to read what the AI wrote.

I'm learning [topic, e.g. React state]. Do not give me the full solution yet.

Problem: [describe what you are trying to do]
My attempt so far:
```
[paste your code]
```

Give me a hint about where my thinking is off, and ask me one question that helps me find the issue myself. Only show the solution if I ask for it.

What to expect: Use this when you are practising for exams or interviews. If you always get the answer immediately, the learning does not stick.

Here is your answer: [paste]. Now act as a sceptical reviewer. List the assumptions it makes, three ways it could be wrong, and how I could check each one quickly.

What to expect: Models can often find flaws in their own earlier answers when asked directly. It is still not proof. Run the code or check the docs.

There are two situations. When you are shipping (a working prototype for yourself, a quick script), leaning on AI is reasonable as long as you test. When you are learning (a course, an exam, interview preparation), copying AI answers skips the struggle that builds the skill. Many students find they can 'build' projects but freeze in an interview when asked to write a loop without AI. Use AI to explain, give feedback and generate practice questions, and write the core code yourself.

College exams and many interviews are done without AI tools. Even in jobs where AI is allowed, you are responsible for what you ship. 'The AI wrote it' is not an answer to a production bug.

Expect questions such as 'How do you use AI in your work?' and 'Tell me about a time AI gave you wrong code'. Strong answers show a process: small tasks, tests, review, verification. A weak answer is 'I ask it and paste the result'. Be ready to solve a problem on a whiteboard or plain editor, and to explain code you wrote with AI line by line.