We tried to write a “normal” roadmap. It immediately split into three contradictory roadmaps and a doodle of a worm wearing a headset. We kept the worm; it felt more honest.
ØVOID is a lab for intentional weirdness.
We’re a two-person experimental software lab building in public uncertainty. We treat AI as a force that turns software into a commodity—so we compete on taste, weirdness, and the courage to ship first drafts.
Enter the lab (mind the loose wires).
Nothing here is final. Some of it is barely behaving.
ST. 01 Mission
AI will flatten software. We build where the flattening can’t reach: taste, risk, and the audacity to ship weird prototypes before they learn to behave.
What we are
ØVOID is a two-person experimental software lab. We build in public uncertainty: small systems, sharp constraints, and prototypes that tell the truth quickly.
If the prototype survives the weekend, it graduates to “maybe.”
What we optimize for
Not maximum polish. Not maximum consensus. We optimize for learning velocity and a particular flavor of legible chaos: enough structure to steer; enough freedom to surprise us.
Constraint: the site is not the product. The lab is the product. The work is the artifact.
ST. 02 Philosophy (manifesto)
We choose experimentation over safe choices, even when the safe choice is very polite. Beautiful chaos is not a phase; it’s an operating system.
AI makes software a commodity—so we make taste the differentiator.
We choose experimentation over safe choices, even when the safe choice is very polite.
Beautiful chaos is not a phase; it’s an operating system.
Co-creation beats control. The future is being written.
We don’t polish the edges off ideas; we see what sharp edges teach us.
If it can’t surprise us, it’s probably not worth building.— Lab notebook margin
ST. 03 Project: AI-native gaming platform
Games created, played, and judged by AI. The judge is synthetic; the drama is real enough.
What “AI-native” means here
Not “a game with AI features.” The platform is built around AI as a collaborator and critic. Players don’t just play a finished game—they witness game rules being proposed, contested, scored, and revised.
Our favorite kind of verdict is the one that reveals the system’s values. Example from an early playtest: “technically illegal but emotionally accurate.” We framed it. Then used it as a design constraint for a week.
- Core loop
- Create → Play → Judge → Mutate. Repeat until the game argues with itself in a way you enjoy.
- Judging style
- Scorecards with categories like audacity, narrative weather, and illegal elegance.
- Lab stance
- Co-creation beats control; first drafts can be rude, and that’s data.
- Non-goal
- Endless feature polish. The safe version is always available later.
ST. 04 Lab notes
We keep receipts—because chaos without documentation is just noise.
We built a “calm mode” for the lab. It lasted 19 minutes before someone added a button labeled “introduce weather.” Calm mode now includes thunderstorms and soft apologies.
Once, we asked the AI to generate a tutorial level. It made an unwinnable arena and titled it “Welcome.” Instead of fixing it, we made “Welcome” a recurring motif: first drafts can be rude, and that’s data.
ST. 05 Signals (the lab admits feelings)
Metrics, but only the kind that help us steer. Some of these are jokes. Some are instrumentation. The best ones are both.
ST. 06 Artifacts
We can’t show external images here. So we label the physical evidence: the stuff that accumulates when you ship fast and listen hard.
A coffee-stained index card labeled “Beautiful Chaos Budget” with three line items: time, courage, snacks.
A tiny JSON snippet: { "mood": "unpredictable", "safety": "experimental" } (crossed out, rewritten twice).
A linty sticker that reads “FIRST DRAFT ONLY” applied to a perfectly fine feature just to keep it humble.
A one-page “Judge’s Almanac” with categories like audacity, narrative weather, and illegal elegance.
A terminal screenshot where the last line is: “Verdict: delightful menace.”
A hand-drawn map of an imaginary arcade called “The Draft Hall,” with exits labeled Ship, Rewind, and Let It Cook.
ST. 07 Contact / entry point
We’re two people. We don’t do noise. If you’ve got a strange problem worth building around, we’ll listen without flinching.
How to approach
Send a link, a demo, or a constraint. We respond best to clear stakes and weird criteria.
- Primary GitHub
- Preferred tone curious, direct, a little mischievous
- Warning label Caution: may contain beautiful chaos.
One last quote (for the road)
Every system tells the truth eventually. Our job is to listen without flinching.— The other one of the two (reading logs)