Building a Game with AI

Why this matters

A block-breaker game feels far more complex than a quiz, yet it is built from the same web languages and a handful of clear ideas. Working with AI on a game shows you how to read and adjust code you didn't fully write, and it shows you where AI helps and where it can't.

The idea

In a block-breaker game, a ball bounces around, breaks blocks, and the player moves a paddle to keep the ball in play. A few core ideas drive it:

• Game loop: the program repeats a cycle many times per second. It moves the ball a little, checks what it hit, and redraws everything. Fast repetition is what makes motion look smooth.
• Collision detection: each cycle the game checks whether the ball has touched a wall, the paddle, or a block, and reacts (bounce, or break the block).
• Score and lives: the game keeps a score that goes up as blocks break, and lives that go down when the ball is missed. When lives reach zero, the game ends.

You can build it with an AI ("Create a simple block breaker game, separate HTML/CSS/JS"), then change it both by re-prompting and by editing values yourself: make the ball bigger, recolour blocks, add more blocks, change speed, show a score, set lives, or make blocks tougher. To edit by hand you reason from variable names and adjust values little by little while watching the result. That is how you understand code that is beyond what you could write from scratch.

This is also where you judge AI honestly. AI can generate working code fast, explain it, and suggest changes. AI cannot guarantee the code is correct (it may hallucinate bugs), know your exact intent, or replace your testing and understanding.

Picture it

flowchart TD
  S[Start: set ball, paddle, blocks, score, lives] --> L[Game loop]
  L --> M[Move the ball]
  M --> C{Collision?}
  C -- Wall/paddle --> B[Bounce]
  C -- Block --> K[Break block, score up]
  C -- Missed --> V[Lose a life]
  B --> R[Redraw]
  K --> R
  V --> Z{Lives left?}
  Z -- Yes --> R
  Z -- No --> G[Game over]
  R --> L

Worked example

You ask the AI to "make the ball change its angle every time it hits the paddle". It rewrites the collision part of the loop. You run it, and it works. Then you want a bigger ball, so instead of re-prompting you find the ball's radius value in the code and increase it, checking the game after each change. When something breaks, you reason about it (or ask the AI), fix it, and test again. The loop of change → run → observe → adjust is the real work. AI is a fast helper inside it, not a replacement for your judgement.

Your turn

With an AI assistant, generate a block-breaker game, then make it yours: change ball size, block colours, speed, add a score and lives. Try editing some changes by hand to practise reading code. Afterwards, note what was difficult, what strategy helped, and one thing AI could and could not do. Then answer the practice questions on the core game ideas.

Recap

• A game runs a game loop many times per second: move, check collisions, redraw.
• It tracks score (up as blocks break) and lives (down when the ball is missed).
• Build and change it with AI, but edit and test yourself to truly understand it.
• AI can write and explain code fast, but it cannot guarantee correctness or replace your testing.