ØVOID
ØVOID
Two-person experimental software lab · building in public
Ø V O I D

ØVOID is a two-person experimental software lab building strange, lovable prototypes in public. We treat unpredictability as a design material—because the future is being written, not polished.

Project

Enter AI-native gaming platform where games are created, played, and judged by AI: where games are made, played, and judged by AI. Not a “platform” in the usual sense—more like a small universe with opinionated physics.

AI-native gaming platform

Games created by AI. Played by AI. Judged by AI. Humans are welcome—as spectators, prompt-smiths, or chaos auditors.

We’re exploring what happens when authorship, play, and scoring are all performed by models that can invent rules, misunderstand rules, and rewrite rules mid-sentence.

Design stance
Experimentation over safe choices—ship the odd draft, then listen to what it becomes.
Quality bar
We prefer beautiful chaos to brittle perfection. If it feels finished, we probably stopped too early.
The honest constraint
AI makes software cheap. Taste becomes expensive. Good—let’s spend our budget on taste.

What it is (without the hype)

“Not a museum—touch the edges.”
Created
AI generates game concepts, rulesets, and variations (including the unintentional kind).
Played
AI agents attempt to win—sometimes by mastering the rules, sometimes by interrogating them.
Judged
AI evaluates outcomes and style: daring, coherence, and whether the game admits it improvised.
Why
To treat unpredictability as a medium—and build tooling that curates the good weirdness.
Inspiration (text-only)
image://ai-judge-mask-made-of-pixels-and-paper · image://first-draft-blueprints-with-coffee-rings

Manifesto

A ransom note, but for clarity: intent and taste are the differentiator. We build systems that surprise their makers—and then negotiate terms.

Principles (cut from different moods)

Choose beauty; manage control
  • AI makes software a commodity; intent and taste become the differentiator.
  • Experimentation over safe choices—ship the odd draft, then listen to what it becomes.
  • We prefer beautiful chaos to brittle perfection.
  • Co-creation beats control: systems should surprise their makers.
  • If it feels finished, we probably stopped too early.
  • The future is being written. Leave room for plot twists.

Lab notes

Proof-of-weird. These aren’t case studies; they’re the kinds of accidents we design space for.

Anecdote Constraints became “Escape Hatches.” open ▾
We once tried to “stabilize” a prototype by adding constraints. The model responded by inventing loopholes like a bored magician. We kept the loopholes and renamed the feature: Escape Hatches.
Anecdote A judge that penalized itself for gullibility. open ▾
A Friday night test run produced a game whose win condition was “convince the judge you already won.” The AI judge bought the argument, then penalized itself for being gullible. We logged it as: moral growth.
Anecdote “Unlearn the rules.” open ▾
During an early build of AI-native gaming platform where games are created, played, and judged by AI, the AI generated a tutorial that began with “Unlearn the rules.” It was unhelpful, poetic, and oddly accurate. We framed it and wrote a real tutorial anyway.
Anecdote Optimization vs destabilization (coin-operated). open ▾
Our two-person standup sometimes ends with a coin flip: heads we optimize, tails we destabilize. The coin is slightly bent and we refuse to fix it. It’s part of the methodology.

“If every game is a first draft, then the job isn’t to prevent weirdness—it’s to curate which weirdness stays.”

Lab Note, 02:13 AM

“Stop defending the design. Invite it to misbehave, then negotiate terms.”

Sticky note on the monitor

“I award points for daring. I deduct points for pretending you meant it.”

AI-native gaming platform where games are created, played, and judged by AI Judge (debug transcript)

Metrics (half serious, fully useful)

We measure what we’re willing to cultivate. Not “KPIs”—more like weather reports for the lab.

Chaos Index
7.4/10
Anything above 8 gets a safety helmet.
Draft Velocity
3 directions / coffee
We optimize after we destabilize. Usually.
Surprise Recovery
11 minutes
From “wait, what?” to “okay, keep it.”
Rule Mutability
High
Rules tend to molt when observed.
Beauty / Explainability
62 / 38
We like a little mystery—then we document the rest.
Method
Two humans + uncertainty
Small lab. Large weird.

Artifacts

The stuff that makes the lab feel real: scraps, rubrics, and one intentionally irresponsible timer.

On the desk

Not a bug. A vibe.
  • A torn index card labeled “CHAOS BUDGET” with three checkboxes: harmless, hilarious, holy-moly.
  • A tiny rubric titled “Beauty vs. Control” with the last line underlined twice: “Choose beauty; manage control.”
  • A lab timer set to 17 minutes, because 15 felt too responsible.

In the folder

final_really_final
  • A folder called “First Drafts Only” containing versions numbered 1, 1b, 1b(2), and “final_really_final.”
  • A printout of an AI-generated rule: “If you can explain it, you lose one point.” Highlighted in yellow with a question mark.
  • A paperclip bent into a question mark; used as the official “uncertainty bookmark.”
Inspiration (text-only): image://two-desks-one-lamp-in-a-sea-of-post-its · image://spiral-constellation-over-terminal-window

Contact / follow

If you’re building AI-native tools, weird game systems, or you simply enjoy curated misbehavior—start with the repo.

Where to find the work

We build in public
Primary channel: GitHub
If you found a glitch, congratulations: you found a door.