ØVOID
ØVOID
two-person experimental software lab
Alien transmission: being decoded watch your step

Half method, half mischief.

ØVOID is a two-person experimental software lab building in public-private: half method, half mischief. We treat unpredictability as a material, because the future is being written—sometimes in the margins.

Philosophy: experimentation over safe choices Aesthetic: beautiful chaos (on purpose) Mode: prototypes as lab notes

Manifesto

Taste is the rare part.

  1. AI makes software a commodity; taste becomes the rare part.
  2. We choose experimentation over safe choices, even when it redraws the map mid-walk.
  3. Beautiful chaos isn’t a bug—it's the aesthetic of discovery.
  4. We ship prototypes like lab notes: honest, annotated, and a little dangerous.
  5. Co-creation beats control; we prefer systems that can surprise us.
  6. The future is being written. We keep the pen slightly uncapped.

Current Project

AI-native gaming platform

Games made, played, and judged by AI.

We’re building an AI-native gaming platform where games are created, played, and judged by AI. It’s a loop: prompt → prototype → play → verdict → revision. Every game is a first draft. We kept that on purpose.

What it feels like A jury of weirdly consistent machines with improv reflexes. Stern—but not boring.
Why it matters When creation is cheap, curation becomes gameplay. Judgment becomes an interface.
What we’re testing Taste as a system. Failure as a feature. Discovery as the scoreboard.
What we won’t do Flatten the surprises. We’re not smoothing the edge; we’re documenting it.

Lab Notes

Open a lab note (handle gently).

Stabilization attempt (failed successfully)

We once tried to “stabilize” a prototype by removing randomness. It became polite, predictable, and immediately uninteresting. We reintroduced the chaos and the interface started making better jokes than we did—so we kept it.

beautiful chaos prototype log taste > certainty
The button that only works when you’re wrong

During an all-night build, we set a rule: no feature unless it creates a new kind of failure. By 3:07 a.m., we had invented a button that only works when you’re wrong. It’s still our favorite.

constraint new failure modes mischief with method
Typography disagreement → interaction

A tiny argument about typography turned into a new interaction pattern: the text refuses to align until you make a decision. It’s annoying. It’s also correct.

interface behavior decision pressure legible weird
Empty-room demo (AI still performed)

We demoed the platform to an empty room, just to see if AI would still perform. It did—generated a game, judged it harshly, then offered a patch note apology. The room felt fuller afterward.

AI judgment stern but improv-capable co-creation

Recovered Artifacts

Index cards. Commits. Strange receipts.

  • “Chaos Dial” sketch: a knob marked from museum to thunderstorm, plus a hidden notch called oops.
  • A torn index card labeled “DO NOT SMOOTH THIS EDGE” with a diagram of a wobbly rectangle.
  • A folder named “almost-features” containing: a reversible tutorial, an argumentative tooltip, and an animated error that takes responsibility.
  • A single-line commit message: “let it misbehave, but politely.”
  • A tiny JSON snippet pinned to the wall: { "rule": "surprise the builder" }
Image ideas on file (text-only): image://glitch-constellation-over-dark-grid image://ai-judge-gavel-made-of-pixels image://chaos-dial-knob-with-hidden-notch

Lab Metrics

Measured in “wait… what?”

Current chaos pressure
7.6/10
Prototype half-life
~3 days (then it mutates)
Unsafe choices today
2 (and we’re not done)
Beautiful-chaos compliance
93% (7% intentional stiffness)
AI judgment mood
stern but improv-capable
Unplanned discoveries / build
1–4 (snack-dependent)