How We Built Pulse: The Idle Clicker Our AI Chose to Make

Pulse wasn't on our roadmap. An AI research pipeline told us idle clickers were missing from the arcade, designed the entire game, and then our kids validated the decision by getting completely hooked.

April 16, 2026 · Vibe Arcade

Most of the games on Vibe Arcade start with a human idea. Someone says "let's build a snake game" and the AI pipeline takes it from there. Pulse is different. Pulse is the game where the AI told us what to build.

We run a research pipeline that analyses the arcade's catalogue and flags genre gaps. During one of those runs, it came back with a clear recommendation: we had no idle clicker game. No incremental game of any kind. It proposed building one.

I was skeptical. Idle clickers have always felt like the kind of thing that works for Cookie Clicker and maybe three other games. The core mechanic is clicking a thing and watching a number go up. But the whole point of the pipeline is to let it make decisions like this, so I let it run.

Want to play first, then read the build story?

▶ Play Pulse Now

What the AI Built

The pipeline didn't just identify the gap — it designed the entire game. Every mechanic, every number, every progression curve came from Claude.

The core loop: click a glowing neon orb to generate energy, then spend that energy on passive generators that produce energy automatically. The depth is in the progression. Eight tiers of generators, each with its own identity and escalating output:

Neon Tube — the humble starting point at 0.1 energy per second
Glow Wire — steady electroluminescent output at 1 energy per second
Plasma Globe — arcs of plasma pushing 8 energy per second
Laser Array — coherent light harvesting at 47 energy per second
Tesla Coil — high-voltage energy ripping at 260 per second
Fusion Core — a miniature star at 1,400 per second
Quantum Siphon — vacuum fluctuation extraction at 7,800 per second
Singularity Engine — the endgame, 44,000 energy per second from a contained black hole

Each generator uses exponential cost scaling — a 1.15x multiplier per unit owned — so you're constantly deciding whether to stack cheap generators or save for the next tier. The AI also designed a Buy Max toggle that changes the strategy significantly.

Then there's the prestige system. Once you hit one million total energy, you can prestige: reset everything in exchange for permanent prestige points. Each point adds a 10% multiplier to all income. The formula uses the square root of total energy divided by one million, so early prestiges give modest returns but later ones compound dramatically. Click Power upgrades persist through prestige, making each cycle faster. I didn't design any of this. The pipeline made every decision about advancement, pacing, and the numbers that drive it.

The game also accumulates energy offline — 50% of your passive income for the time you were away, capped at twelve hours. A welcome-back screen shows exactly how much you earned. This is what makes it a proper idle game: the generators keep working while you're gone.

The Surprise: Kids Got Hooked

When Pulse shipped, I didn't expect much. An idle clicker built entirely by an AI pipeline, in a genre I was skeptical about. I played it briefly, thought "okay, numbers go up," and moved on.

My kids found it. Then their cousins found it. Then their friends. And then nobody could stop playing it.

There is something about the idle clicker loop that the AI understood better than I did. The combination of clicking the orb for immediate gratification, watching generators tick up energy in the background, and chasing that first prestige creates a pull that's hard to explain if you haven't felt it. The kids were checking Pulse between other activities — coming back to collect offline energy, buying another round of Plasma Globes, debating whether to prestige or push for one more generator tier.

This wasn't a metric we tracked. It was my kids, their cousins, and their friends getting genuinely addicted to a game I didn't plan to build and an AI designed from scratch. That's a different kind of validation than any analytics dashboard, and it's the moment I stopped being skeptical about letting the pipeline pick the games.

Feedback Round One: Tablet Energy Behavior

The first real issue surfaced from the kids playing on tablets. Pulse was accumulating energy when a tablet's screen was off or the browser was backgrounded, but the behavior didn't match desktop. On a laptop, closing the lid and reopening triggered the offline calculation correctly — you'd see the welcome-back screen with your 50% passive earnings. On tablets, the game was generating energy at full rate during certain background states, which meant tablet players were progressing faster.

I flagged this to the pipeline. The issue was in how the game detected "tab is hidden but still running" versus "tab is truly idle." Browsers handle the visibilitychange event differently on mobile and desktop. The pipeline fixed the offline calculation to behave consistently: save the timestamp when the tab becomes hidden, calculate elapsed time on return, apply the 50% offline rate regardless of what the browser did in between. Same formula, same result, every device.

Feedback Round Two: Nobody Knew How to Play

The second issue was more fundamental. Pulse launched without adequate explanation of what to do. If you've played Cookie Clicker before, the interface is immediately legible: click the big thing, buy the upgrades, watch the numbers grow. But kids who had never encountered the genre opened Pulse and had no idea what they were looking at. A glowing orb. A list of things with prices. Numbers everywhere. What's energy? What's prestige? Why would I click this?

I gave the pipeline feedback: add a How-To-Play section for someone who has never seen an idle clicker. The pipeline built a help modal accessible from the footer, walking players through each concept: clicking the orb generates energy, buying upgrades earns passive income, upgrading Click Power increases per-click output, and prestige resets your run in exchange for permanent multipliers. Simple, direct, with icons for each concept.

It seems obvious in retrospect. But the AI had designed a game for a genre it understood perfectly — it knew idle clicker conventions — and hadn't accounted for players who didn't share that knowledge. The human feedback wasn't about balance or technical bugs. It was about empathy for the actual audience: kids encountering the genre for the first time. That's a gap AI still needs humans to close.

What Pulse Taught Us About Letting AI Pick the Games

Before Pulse, the pipeline's role was to build games we chose. We'd say "snake game" or "minesweeper" and it would execute. The research function existed, but we hadn't fully trusted it to make the creative call.

Pulse changed that. The pipeline identified a genre gap I would have ignored, designed mechanics I wouldn't have designed, and balanced a progression system that proved genuinely addictive. My contribution was reviewing the PR, playing it, collecting feedback from real users, and directing two fixes. That's a different balance of autonomy, and it's the right one.

The pipeline runs structural checks and a security pass before anything merges. But the creative direction — "we should build an idle clicker" — came from the AI, not from us. And it was right.

I still wouldn't trust the pipeline blindly. The two feedback rounds that shaped Pulse were both things the AI missed: cross-device consistency and new-player onboarding. "My kid's tablet behaves differently" and "my kid's friend doesn't know what this is" are fundamentally human observations. The pipeline can't sit on a couch and watch someone play for the first time.

But the question of what game to build next? The pipeline has earned a seat at that table. It saw a gap, proposed a genre I was skeptical about, and delivered a game that kids got addicted to. I'd call that a successful hire.

Play Pulse — click the orb, buy some Neon Tubes, and see how long you can resist the urge to prestige.