ØVOID
ØVOID two people • experimental software lab • building in public view
Unstable, on purpose
GitHub Note: This page may rearrange itself.
Enter the lab (mind the loose ideas). ØVOID / field-notes

Beautiful chaos,
built deliberately.

ØVOID is a two-person experimental software lab building in public view: unstable on purpose, beautiful by design. We treat AI as the force that turns software into a commodity—so the only durable advantage is taste, velocity, and brave experiments.

Meet our current project Read the philosophy Browse the lab notes

Choose your route (it’s all one page; the order is the experiment).

Think of this like a choose-your-own-adventure page: each chapter is a door. Open whichever feels most “you,” then loop back for the rest.

Door A: The platform (games made, played, judged by AI) Turn to §1 — the current build
PAGE 01
Door B: The philosophy (taste, velocity, brave experiments) Turn to §2 — why we do it this way
PAGE 02
Door C: The artifacts (rules (temporary), logs, metrics) Turn to §3 — evidence of life
PAGE 03

§1 — Current Project: an AI-native gaming platform

We’re building a platform where games are created, played, and judged by AI. Not a tech demo: a playground where the rules can be invented, contested, and evaluated—at runtime—by systems designed to disagree interestingly.

Project codename REFEREE//LOOP
Loop make → play → judge
Constraint legible weirdness
1) Creation

A maker-model proposes rules, objectives, and win conditions—sometimes conventional, sometimes delightfully wrong. The point is to generate playable novelty, not endless text.

2) Play

Player-models test the game. They exploit loopholes, discover degenerate strategies, and do the thing humans do best: take rules literally.

3) Judgment

A referee-model issues verdicts and ratings. If a game is confusing, the referee should say so—then explain why in plain language.

What this enables

  • Games that evolve faster than “content pipelines.”
  • Rules that can be interpreted (and argued) by judges.
  • Replayability driven by variation, not grind.
  • Proof that evaluation is part of creation—not an afterthought.
Gameplay log Game #0412
“The Door That Judges You Back.” Rating: 9/10 for audacity, 2/10 for doors.
— Gameplay Log, REFEREE//LOOP

What we’re optimizing for (not the usual)

AI is making the how cheap. So we obsess over the why—and the exact shade of weird in the middle. The platform is designed to surface taste: in the prompts, in the constraints, in the scoring, in the refusal to sand down every edge.

Turn to §2 (philosophy) Turn to §3 (lab notes) Back to the doors

§2 — Philosophy: experimentation over safe choices

Our default stance is: ship the first draft, then ship the draft of the first draft. Beautiful chaos isn’t a bug; it’s a creative environment—with guardrails so it stays legible.

Team size two people
Method build in public view
North star interesting & durable
Ø

AI makes software a commodity—so we spend our time on weirdness, taste, and acceleration.

Ø

We choose experimentation over safe choices, even when the safe choice is begging politely.

Ø

We ship first drafts, then we ship the draft of the first draft.

Ø

Beautiful chaos is not a bug; it’s a creative environment.

Ø

Co-creation beats prediction. The future is being written.

Ø

If it’s predictable, it’s probably already been done.

Internal motto working rule
Make it strange. Make it legible. Let it evolve.
— Internal Motto

How “legible chaos” works here

  • Intentional constraints: every experiment has a scope and a stop condition.
  • Readable outputs: if an AI judges a game, it must explain itself.
  • Versioned weirdness: we keep fossils (logs) and compare them over time.
  • One rule of taste: if it’s dull, it doesn’t ship.
Lab note whiteboard
Stop asking if it’s shippable. Ask if it’s interesting enough to survive shipping.
— Lab Note (whiteboard)
Turn to §1 (project) Turn to §3 (lab notes) Back to the doors

§3 — Lab Notes: artifacts, metrics, and proof of life

We keep receipts. Not to sound serious—just to stay honest. Below are snapshots of the lab: temporary rules, coffee-stained decisions, and measurements we pretend are scientific.

Archive field-notes
Evidence type logs / drafts / verdicts
Warning odd on purpose

Artifacts (found in the lab)

  • RULES (temporary) label: chaos-garden-flowchart-vines

    A torn sheet labeled “RULES (temporary)” with three bullet points and one doodle of a crown made of brackets.

  • first-drafts-only/ label: first-draft-game-cartridge-melt

    A folder named “first-drafts-only/” containing exactly one file: README_WIP_FINAL_FINAL2.txt.

  • AI verdict label: ai-referee-whistle-made-of-pixels

    A small printout of an AI judge’s verdict: “Technically beautiful. Emotionally confusing. Approved.”

  • Index card label: two-desks-one-lab-schematic

    A coffee-stained index card: “When in doubt, add a referee.” On the back: “When in more doubt, let the referee argue with itself.”

  • Changelog label: neon-circuit-bloom-on-black

    A mini changelog entry: “v0.0.0 — Introduced chaos. v0.0.1 — Admitted it was intentional.”

Chaos Budget
38% (soft cap, frequently ignored).
First-Draft Velocity
1.6 ideas/hour (measured in scribbles, not commits).
AI Surprise Index
“high” (calibrated by eyebrow distance).
Judgment Latency (REFEREE//LOOP)
2–7 seconds, depending on how philosophical the referee feels.
Beautiful-Noise Ratio
3:1 when caffeinated; 1:2 near deadlines.
Anecdote delightful accidents
We once tried to write a “clean” roadmap. It looked responsible for six minutes, then we added a column called “delightful accidents” and it immediately became the only column with dates.
— from the desk (near the coffee rings)

What we are (plainly)

ØVOID is a two-person experimental software lab. We build software like sketches: fast, honest, revisable. If you saw something odd, you probably saw the point.

Between sips of cold tea why
If AI commoditizes the how, we obsess over the why—and the exact shade of weird in the middle.
— ØVOID
Turn to §1 (project) Turn to §2 (philosophy) Back to the doors