ØVOID
ØVOID
two-person experimental software lab • building loud prototypes & public-facing whispers
Enter the lab (watch your step). Nothing here is finished. That’s the point.

We build systems that learn to surprise us—then we keep them legible.

ØVOID is a two-person experimental software lab building in public-facing whispers and loud prototypes. We treat unpredictability as a design material—because the future is being written, not announced.

Browse experiments, not promises. Try ↓ for sections

See what the platform is becoming GitHub

Mission

a concrete monument to soft uncertainty
Summary
ØVOID is a two-person experimental software lab building in public-facing whispers and loud prototypes. We treat unpredictability as a design material—because the future is being written, not announced.
  • Experimentation over safe choices, even when it looks like beautiful chaos.
  • Co-creation isn’t a feature; it’s the default setting of reality.
  • First drafts are deliberate—iteration is the craft.

What we do

build → test → reframe

We’re building an AI-native gaming platform where games are created, played, and judged by AI. Humans set the tone, aim, and constraints; the system supplies relentless iteration and surprising interpretations.

Why “judged”?
Because scoring systems shape behavior. We don’t want a perfect system. We want a system that keeps producing interesting mistakes— and can explain why it rewarded something, even when it’s wrong.
How it feels
Like an arcade cabinet built from blueprints and doodles: strict edges, soft outcomes. A game can “work” and still surprise you.

Pardon the beautiful mess—fresh paint is still thinking.

Platform details GitHub

Current project: an AI-native gaming platform

Create → Play → Judge → Laugh → Repeat

We’re exploring what happens when the platform itself becomes a collaborator: the game designer, the player, and the referee are all machine processes—with tastes we can shape and audit.

The loop
Create
An AI drafts rules, levels, and interactions under constraints you set (style, difficulty, weirdness, clarity).
Play
AIs (and humans, when invited) run sessions at scale, searching for edge cases and emergent tactics.
Judge
Judge agents score outcomes—sometimes for “fairness,” sometimes for “audacity,” always with rationale you can inspect.
Field story (from the lab)
At 2:13 a.m., we argued whether “fair” is a rule or a vibe. By 2:40, we’d built a tiny judge that rewarded “audacity” and immediately regretted it—in the best way.

Telemetry we actually respect

not KPIs • signals
Prototype Half-Life
3 coffees (±1 existential sigh)
If it survives longer, we’re either onto something—or procrastinating.
Chaos Coefficient
0.72
When above 0.80, we add a tiny rule. When below 0.60, we remove one.
Polish Budget
15% shine • 85% surprise
Clarity stays. Cosmetics are negotiable.
Judge Mood Variance
spiky, with delightful outliers
The point isn’t stability. The point is explainable weirdness.

Tip: hover/tap the meters to “stress test” the reading.

Reference wall (text labels only)
  • image://glitched-arcade-cabinet-blueprints
  • image://ai-judge-scorecards-with-doodles
  • image://chaos-spiral-turning-into-a-grid
  • image://two-desk-lab-at-midnight-with-post-its
  • image://neon-breadcrumbs-on-black

Philosophy

taste • nerve • iteration
  • AI makes software a commodity; taste, nerve, and iteration become the differentiator.
  • We choose experimentation over safe choices, even when it looks like beautiful chaos.
  • We ship first drafts on purpose—then watch what emerges when systems talk back.
  • Co-creation isn’t a feature; it’s the default setting of reality.
  • If it feels too polished, we sand it down until it learns to surprise us.
Practical stance
We don’t chase “perfect.” We chase a tight loop: build a thing, give it a judge, let it misbehave, then tighten the language around it so the weirdness stays intentional.

Quote wall

notes, scribbles, kettle wisdom

“If AI makes software cheap, we spend our budget on weirdness and clarity.”

— Lab Note, taped above the kettle

“We don’t want a perfect system. We want a system that keeps producing interesting mistakes.”

— One of the two humans at ØVOID

“A first draft can be a place to live—just don’t pretend it’s a fortress.”

— Margin scribble from a design sketch

“We optimize for ‘huh, I didn’t expect that’ without optimizing for ‘what even is this?’”

— The other human at ØVOID

Artifacts (the lab leaves paper trails)

evidence of beautiful chaos
A-01torn index card

“Rules are suggestions until they’re tested.”

A-02mood.json

A tiny JSON snippet with only two keys: "brave" and "feral".

A-03leaderboard mock

A screenshot where the top rank is “Mostly Harmless.”

A-04changelog draft

“Made the judges more curious. Side effect: they now ask questions.”

A-05post-it diagram

Create → Play → Judge → Laugh → Repeat.

A-06paperclip bundle

Two printed pages: one titled “Order,” the other “Chaos,” both highlighted in the same color.

Lab note (live excerpt)

reshuffle for a different angle

At 2:13 a.m., we argued whether “fair” is a rule or a vibe. By 2:40, we’d built a tiny judge that rewarded “audacity” and immediately regretted it—in the best way.

tag: judge-design status: draft, intentionally

Browse experiments, not promises.

Contact surface

minimal, on purpose

We’re ØVOID. Two people. A lot of prototypes. If you’re building in adjacent weirdness—AI evaluation, procedural design, emergent play—start with the code trail.

See the work: repos over rhetoric

GitHub Back to top
Sticky notes (mood swings)
We have a sticky note that says “Chaos is a feature” and another that says “Chaos is a bug.” We move them around the monitor depending on the day.
(Reserved for required injected content.)