Vibe Coding · 课时

发现错误与糟糕模式

人工智能常见的错误，以及如何发现这些错误。

第 3 / 4 课13 个步骤

发现错误与糟糕模式 是 CoddyKit 上的免费 Vibe Coding 课时。这是第 3 节课，共 4 节。你可以在下方免费阅读本课时的完整内容 — 然后在浏览器中使用内置代码编辑器和全天候 AI 导师进行实践。这是 Vibe Coding 学习路径的一部分，你的进度在网页和 CoddyKit 应用中同步。 Vibe Coding 课程共包含 4 节课。

本课时的部分内容尚未翻译，以英文显示。

Learn the AI Mistake Patterns

AI makes the same kinds of mistakes over and over. Once you know the patterns, you'll spot them in seconds — like a detective who's seen the same trick a hundred times.

This lesson is a field guide to the most common bugs and bad patterns in AI-generated code, plus how to catch each one. No deep CS background needed.

You don't memorize fixes — you learn what to watch for.

Pattern 1: Missing Edge Cases

AI loves the 'happy path' and often forgets the messy reality: empty lists, missing data, zero, or null.

Run this average function with a normal list — it works. But picture what happens if the list is empty.

function average(numbers) {
  let sum = 0;
  for (const n of numbers) sum += n;
  return sum / numbers.length;
}

console.log(average([10, 20, 30])); // 20, fine
console.log(average([]));            // NaN — divide by zero!

Catching Edge Cases

The empty list gave NaN (not a number) because it divided by zero. In a real app that could show 'NaN' to a user or crash later.

The catch: whenever you see code that loops, divides, or reads a property, ask 'what if it's empty or missing?'

Fix prompt you can paste: 'Handle the empty-list case and any null inputs in this function. Return a sensible default and explain your choice.'

Pattern 2: Reading Missing Data

AI often assumes data is always shaped perfectly. Reach into an object that might not have what you expect, and you get the classic crash.

This one throws a real error. Run it and read the message.

function getCity(user) {
  return user.address.city; // assumes address always exists
}

const u = { name: 'Sam' }; // no address!
console.log(getCity(u));    // TypeError: Cannot read properties of undefined

Catching Missing-Data Bugs

That Cannot read properties of undefined error is one of the most common in all of web development. AI generates it constantly because it assumes ideal data.

Watch for chains like a.b.c where any middle piece could be missing. The safe pattern uses optional chaining: user.address?.city, which returns undefined instead of crashing.

When you see deep property access on user input or API data, flag it for review.

Pattern 3: Hardcoded Secrets

A dangerous favorite: AI pastes an API key or password directly into the code to make the example 'just work.'

If this ships to GitHub or the browser, anyone can steal your key and run up your bill. Never let a literal secret survive review.

// DANGER: secret key hardcoded in the source
const STRIPE_KEY = 'sk_live_51HxAbC9secretrealkey';

fetch('https://api.stripe.com/charge', {
  headers: { Authorization: 'Bearer ' + STRIPE_KEY }
});
// Keys belong in environment variables, never in code

Pattern 4: Silent Failures

Sometimes AI 'handles' an error by swallowing it — the app keeps going as if nothing happened, hiding the real problem.

The empty catch below means if the fetch fails, you get no warning, no log, just mysteriously missing data.

async function loadData(url) {
  try {
    const res = await fetch(url);
    return await res.json();
  } catch (e) {
    // swallows the error silently — bad!
    return null;
  }
}
// You'll never know WHY it returned null

Pattern 5: Overcomplicated Code

AI sometimes writes 30 lines for a 3-line job, or adds clever abstractions you didn't ask for. Complexity is where bugs hide and future-you gets lost.

If a simple task produced a tangled result, push back. Try: 'This feels more complex than the task needs. Rewrite it as simply as possible and explain why each part is necessary.'

Simple code you understand beats clever code you don't.

Pattern 6: It Doesn't Match the Goal

The sneakiest bug isn't broken code — it's working code that solves the wrong problem.

You asked to sort newest-first; AI sorted oldest-first. You asked to email active users; it emailed everyone. No error, just the wrong outcome.

Always re-read your original ask and check the output against your intent, not against 'does it run.' This is the bug AI can't catch for you.

Let AI Hunt Its Own Bugs

You don't have to find everything alone. Point AI at the code with a sharp, suspicious prompt and it'll surface many of these patterns.

Keep this prompt handy — it targets exactly the patterns from this lesson.

Audit this code for bugs and bad patterns. Check specifically for:

- Missing edge cases (empty, null, zero)
- Crashes from missing data (deep property access)
- Hardcoded secrets or API keys
- Silently swallowed errors
- Overcomplicated logic
- Anything that doesn't match this goal: [describe your goal]

List each issue, the line, and a fix.

Build Your Spotting Instinct

Your bug radar, summarized:

Loops and division? Check empty and zero.
Deep property access? Check missing data.
Keys or passwords? Never hardcoded.
Empty catch blocks? Silent failure.
Too much code? Ask for simpler.
Always: does it match what I actually asked for?

Run this scan on every chunk and you'll catch the vast majority of AI mistakes.

Quick Check

AI writes return user.address.city; and some of your users have no address. What kind of bug is this, and how do you catch it during review?

在交互式课程中回答

Recap: Catch the Common Bugs

The AI mistake patterns you can now spot:

Missing edge cases — empty lists, null, divide-by-zero.
Missing-data crashes — deep property access on imperfect data.
Hardcoded secrets — keys must live in environment variables.
Silent failures — empty catch blocks that hide errors.
Overcomplication — push for the simplest version.
Wrong goal — working code that solves the wrong problem.

Scan for these every time, and let AI audit itself too. Next: deciding what to keep, fix, or reject.

免费开始

用 AI 导师学习 JavaScript — 免费

在浏览器中编写并运行真实代码，获得全天候 AI 导师的即时帮助，并在网页或应用中继续学习。

课程: 25
课程: 100

免费开始学习查看完整课程

常见问题解答

「发现错误与糟糕模式」课时是免费的吗？

是的 — 「发现错误与糟糕模式」的完整文本可在网页上免费阅读。要进行交互式练习（内置代码编辑器和全天候 AI 导师）并解锁 Vibe Coding 课程的其余内容，请升级到 CoddyKit PRO。 Vibe Coding 课程共包含 4 节课。

「发现错误与糟糕模式」这节课中我会学到什么？

人工智能常见的错误，以及如何发现这些错误。你通过在浏览器中直接运行的动手代码来练习 Vibe Coding，全天候 AI 导师会在你学习这节课的过程中回答你的问题。

学习 Vibe Coding 需要有经验吗？

无需任何先前经验。CoddyKit 上的 Vibe Coding 课程适合初学者到高级学习者，你可以从这里开始或从头开始，按照自己的节奏学习。这是第 3 节课，共 4 节。

「发现错误与糟糕模式」课时需要多长时间？

大多数 CoddyKit 课程大约需要 5–10 分钟。每节课都很精短且互动，所以你能稳步进步，并在网页和应用中从离开的地方继续。

我能在这节 Vibe Coding 课中编写并运行代码吗？

能。每节 Vibe Coding 课都包含内置代码编辑器，你可以在浏览器中直接编写并运行真实代码，并获得即时 AI 反馈 — 无需本地设置。

此课程中的所有课时

← 返回 Vibe Coding