My case: 40 game ideas → a playable portfolio, solo — part 1, the vibecoding assembly line

The vibecoding thesis says building stopped being the bottleneck. Here’s me testing that claim at its most extreme: a one-person game studio funnel — ~40 ideas collected, about half shortlisted, playable prototypes built in bulk, the strongest promoted to full production. This is the live log.

The funnel (idea → MVP → production)

1. Collect without filtering. ~40 ideas captured raw, in my own words — arcade loops, runners, puzzle mechanics. Ideas are graded, never discarded.
2. Shortlist by buildability × fun-potential. About half made the cut for prototyping.
3. MVP wave. Each shortlisted idea becomes a self-contained HTML5 canvas prototype: one file, one core mechanic, one input axis, playable on desktop and phone. AI builds from a tight design brief; I playtest and tune the feel — the part no model judges for you.
4. Promote the winners to production. The strongest few get the full treatment: in-gameplay tutorials, difficulty curves and biomes, “juice” (slow-mo deaths, combo toasts, screen shake), balancing passes, original soundtracks — including music from my own AI-music label case, which is exactly the kind of cross-project leverage a portfolio makes possible.

Current production wave includes titles like BublDog (a French bulldog on a trampoline popping bubbles) and TwoBallz: Super Rush (a two-element runner where fire and ice swap to break what blocks you) — small, finished-feeling arcade games, not tech demos.

Distribution: web-first, stores only for winners

• Now: HTML5 → my own web playground for testing → browser-game platforms (itch.io; portals like Poki and CrazyGames that bring their own audience and share ad revenue). Publishing costs nothing; someone else owns the traffic.
• Later, winners only: wrap for Google Play with Capacitor, monetize with ads. Store effort is expensive — it’s spent only on games the web already proved.

The sequence is the whole strategy: prove fun where publishing is free; pay the store tax only on winners.

The log — results as they land

• 2026-08 (part 1, this post): ~40 ideas collected and graded; ~20 shortlisted; MVP wave built (12 in half a day at peak); production treatment underway on the lead titles (tutorials, biomes, juice, balancing, original soundtracks); test playground live, portal submissions next. Revenue so far: none — by design at this stage. Next checkpoints: first portal publications, first play metrics.

Updates get appended here with dates as real numbers arrive — wins and failures both.

Part of the personal log. The thesis this tests: vibecoding for solopreneurs; the sibling experiment: the AI music label case.