How We Built Super Brick Breaker With AI — Power-Ups That Hurt, Edges That Warp

April 14, 2026 · By Vibe Arcade · 7 min read

The brief for Super Brick Breaker was about as simple as they get: a breakout-style game with a neon look, power-ups, and levels that feel different each time. The AI returned all of that — plus a power-up system with eleven entries, two of which actively hurt the player, and teleporter strips along the edges of the playfield that warp the ball — and your paddle — straight through to the other side. We kept both.

Play Super Brick Breaker →

Phase 1: The Brief

PROMPT

The starting prompt described a classic breakout game: paddle at the bottom, ball bouncing, walls of bricks to clear. Neon visual aesthetic. Some power-ups. Procedurally generated levels so no two runs are identical. Three lives. Simple.

Breakout is one of the most imitated game templates in existence, which made this an interesting test. The mechanics are so well-understood that the AI could simply replicate the standard without thinking about it — or it could use that foundation to build something with more texture. We weren't sure which we'd get.

Phase 2: What Came Back

The first working version had the core mechanics working cleanly: paddle physics, ball reflection angles based on impact position, brick hit detection, and a three-lives system. What was unexpected was the power-up system — not just that there were eleven of them, but how they were designed.

Most breakout games include power-ups that help the player. Wider paddles, extra balls, laser cannons — all positive. Super Brick Breaker's list included those, but also two that actively hurt you:

Shrink — makes the paddle narrower. If you catch it, the ball becomes harder to control. The visual design signals the hazard (it drops in red, marked with inward-pointing arrows and a danger X), but in a fast session, you might catch it by accident. The AI designed a power-up system with genuine downside risk, not just a list of buffs.
Mirror — inverts your left and right inputs, so the paddle runs the opposite way from the key you are holding. Not damaging directly, but disorienting. Until you lose a life or catch a Reset, every correction you make comes out backwards.

Including hazardous power-ups was not a direction we gave. The AI decided that a more interesting power-up economy was one where not every drop is worth catching — where you have to pay attention and make decisions about what to collect. That's a better design than eleven unconditional buffs, and it wasn't in the brief.

Phase 3: Teleporter Strips

The teleporters appeared in the first level design iteration, and they are not bricks — they are strips along the edges of the playfield. A top strip turns the ceiling into a portal: a ball that would have bounced off the top of the screen reappears at the bottom instead, still travelling upward, with the whole board to cross again. Side strips wrap the paddle around the left and right walls, so sliding off one edge brings you back in from the other. The AI included them without being asked.

Our initial reaction was skepticism. Breakout's satisfaction comes from reading the ball's trajectory and positioning accordingly. Teleporters break the straight line between where the ball is and where it is going, which disrupts that skill loop. We spent some time testing whether they made the game worse.

They didn't. The key was that teleporters are staged and visible — they don't appear at all until level eight, and when they do they're drawn as animated strips on the edge of the playfield rather than hidden inside the wall. The effect is surprising the first time, then anticipated and even used strategically. Paddle stranded on the far side of the screen with the ball dropping at the near edge? With side strips active, sliding off the near wall is quicker than running the full width back. The chaos is bounded and navigable.

We kept them.

Phase 4: The Safety Net

The Safety Net — a one-shot catch net at the bottom of the screen that bounces the ball back instead of letting it fall through — emerged naturally from the power-up design pass. It's the most straightforwardly helpful item in the game: catch it and your next near-miss won't cost a life. For early levels, it's a training wheel. For later levels, when the layouts are dense and indestructible steel blocks close off the easy angles, it's a meaningful reprieve.

The interesting design choice was making it a power-up rather than a permanent feature. A permanent Safety Net would change the risk profile of every session — you'd never be in danger of losing a ball. As a single-use drop, it creates moments of relief without removing the tension that makes the game interesting. The AI understood the difference.

Phase 5: Thirty Hand-Designed Levels

The brief asked for levels that feel different each time. What shipped isn't procedural at all: thirty hand-designed layouts, each one authored as a text grid — solid walls, pyramids, checkerboards, fortresses, diamonds, zigzag ribbons, spiral arms, letter shapes, rings, hourglasses. What varies from run to run is what falls out of them. Levels 1–2 seed three to five power-ups to help players get oriented; from level six on it drops to one to three, so the help thins out as the layouts get harder. Clear all thirty and the game loops into Endless mode, recycling the layouts with faster balls and a rising score multiplier.

That's the opposite of what the brief asked for, and it's the better answer. A generator that produces a fresh grid every run also produces grids nobody has vetted — some trivial, some barely completable. Fixed layouts can be tuned one at a time, and the introductions can be placed: armored bricks arrive at level four, steel and the top teleporter at level eight. The randomness that survives is in the drops, where a bad roll costs you a power-up rather than a run.

What We Learned

Hazardous power-ups are a design decision, not a default. Most game generators default to buffs. The AI's decision to include Shrink and Mirror came from treating the power-up system as an economy with risk and reward rather than a list of bonuses. When the AI makes a non-obvious design choice that improves the game, it's worth understanding why — so you can recognize when it's doing it again.
Chaos with structure is interesting; chaos without structure is frustrating. Teleporter strips work because they arrive late, they're drawn right on the edge of the playfield, and their behavior is consistent once you learn it. The unpredictability is bounded. That's the difference between a mechanic that adds tension and one that just annoys.
Randomness needs bounds, not free rein. The variety in a run comes from the drop table — which power-ups appear, and where — and that randomness is constrained by level and by difficulty: the first two levels seed the most drops, and Hard thins the pool so you have to earn the good ones. The layouts stay fixed, which is what makes the curve tunable at all.
Test on the power-up you hate. Shrink made us defensive. Our instinct was to remove it because it felt punishing. Testing proved it was working correctly: it created decision-making around catch vs. avoid, which added depth. Before removing something that feels bad, verify whether it's bad or just unexpected.

The Result

Super Brick Breaker is a breakout game where the power-up drops are decisions, not freebies. The teleporter strips turn the edges of the playfield into part of the board. Thirty hand-designed layouts and a randomized drop table mean no two sessions play out the same. A brief that was deliberately simple produced a game with more strategic texture than we expected.

Play Super Brick Breaker — and see how many levels you can clear.