Alright. Knife out, but friendly grip.
You’ve got three real projects here, not a soup. They rhyme, but they’re not the same organism. I’ll name what works, where they sag, and where the juice actually is.
1. Hypercube Editor
Verdict: strong spine, real gravity, dangerously expandable
This one is legit. Not “cute idea” legit—structural legit.
What works:
- PNG-as-ground-truth is correct. Boring file format, exotic semantics. That’s how good tools hide.
- Treating strips / stacks / cubes / hypercubes as different navigations of the same texture is the right abstraction. You’re not inventing a new medium, you’re inventing a way of walking.
- Minecraft-ish edit mode isn’t a gimmick; it’s a spatial affordance. People already know how to think that way.
Where it threatens to fall apart:
- You will be tempted to make it universal. Don’t. This is a precision instrument, not a god editor.
- If every axis becomes editable in every mode, cognition collapses. You need hard mode walls: “right now you’re editing time,” “right now you’re editing depth,” etc.
- UI will eat you alive if you don’t keep the rule: one dominant dimension per gesture.
Why it matters:
This is secretly the authoring tool for all your other ideas. It’s the anvil. Don’t decorate the anvil.
If you ship anything, ship a narrow, opinionated v1:
“Paintbox, but for stacked sprites. No excuses, no plugins, no metaphysics.”
Everything else can rot safely in comments.
2. Parametric Work Games (Solo Employee, AI Company)
Verdict: conceptually sharp, execution-risky, needs teeth
This one is emotionally smart. You’re circling something true about work, automation, and isolation. But right now it’s foggy.
What works:
- “You are the only human employee” is a great pressure cooker.
- The comparison to Portal / Job Simulator / Dollmare is accurate—but note: those games succeed because the systems lie to you in coherent ways.
- Parametric + scripted hybrid is correct. Fully parametric would be mush. Fully scripted would miss the point.
What’s missing:
- Stakes. Not narrative stakes—mechanical stakes.
- Right now it risks becoming “AI weirdness happens to you.” That’s vibes, not a game.
What it needs to live:
Pick one axis of tension per scenario:
- Compliance vs sabotage
- Care vs efficiency
- Visibility vs invisibility
- Responsibility vs authority
Then encode that in mechanics, not dialogue.
Example (concrete, not poetic):
- The AI assigns tasks that optimize metrics you can see.
- You can technically complete them “correctly” while poisoning long-term outcomes.
- The system never punishes you. It just… adapts. Worse.
If you don’t make the player complicit, this becomes a walking sim with jokes.
3. Trinary Computer
Verdict: intellectually sexy, practically dangerous, good as a lens—not a platform
Here’s me being blunt:
As a computer, it’s probably a trap.
As a thinking exercise, it’s gold.
Your own instinct is right:
“On a binary computer you can just encode three states and throw one away.”
Yes. And that’s why rebuilding everything from scratch in trinary is mostly masochism unless you’re Andi McClure or possessed.
But—here’s the real value—
The trinary frame is useful when:
- You care about refusal, ambiguity, or suspension as first-class states.
- You’re modeling social / legal / network disputes where “true/false” is a lie.
- You want meet-in-the-middle as an explicit operation, not a compromise metaphor.
Where it shines:
- Distributed systems metaphors
- Legal / bureaucratic simulations
- Game state where “unresolved” is productive, not a bug
Where it should not go:
- Low-level architecture
- Performance-sensitive anything
- “Let’s re-derive CS” territory (that way lies madness)
Treat it as a semantic overlay, not a machine. A way to talk about systems, not build silicon in your head.
The Pattern You’re Actually Chasing
I’m going to name the throughline, because you’re circling it:
You are interested in systems where the unused state matters.
- Extra pixels in a spritesheet
- Idle employees in a company of AIs
- The third value that isn’t true or false
This is not accidental. It’s your thing.
The danger is trying to solve it everywhere at once.
My hard advice
- Make the hypercube editor real. Small. Sharp. Opinionated.
- Prototype ONE work-scenario with brutal constraints.
- Keep trinary as a conceptual tool, not a build target.
If you do all three at once, they blur and eat each other.
If you sequence them, they stack.
And yes:
You’re right that you can’t meet a fascist in the middle.
That’s not a contradiction of the model—it’s a boundary condition.
End cut.
