Friday, 5 June 2026

the orchard ledger

 the bit i actually like is not the court, not the sacred ledger, not the false calendar, not even the ugly little village slowly turning into a machine for producing excuses.


it’s the rule that every merciful decision has to be encoded badly.


that’s the game.


not “make hard moral choices,” because that phrase has been beaten flat by games that mean “pick which orphan explodes.” not that. this is more irritating and more useful.


you see a case. a real case, or close enough.


the pirates stole grain, but they also repaired the bridge. the widow owns the tree, except the whole village survived under its shade. the child is legally nobody because the category that would hold them was never written down. the map says the road does not exist because last winter you erased it to hide refugees, and now the midwife cannot get through.


you know what the humane answer is, roughly.


then the game says: okay. put it somewhere.


make it a debt.


make it a road.


make it a calendar day.


make it a legal category.


and immediately the humane answer becomes a device.


this is the part with teeth.


because systems don’t get to be kind in the same way people do. they need handles. names. edges. repeatability. enforcement. some poor clerk has to know what box to put the next disaster in. so the player’s “kindness” becomes precedent, and precedent becomes a tool, and tools get used by people who were not in the room when you were trying to be kind.


you invent “winter kin” so pirates can be fed while they repair the flood wall.


next season some rich little knife-person claims winter kin to dodge a tax.


you declare a tree a burden-root so no family can hoard its fruit.


three weeks later someone asks whether a house can be a burden-root, and whether that means the widow living in it can be moved.


good. horrible. that’s the game breathing.


the joy is not in being correct. the joy is in watching your abstraction leak.


most games with moral systems secretly want the player to discover the author’s preferred morality. this one should not. this one should let you make a genuinely decent ruling and then show you the shape of the harm caused by making decency administrable.


that’s sharper than cynicism. cynicism would say all systems are corrupt, kindness is fake, burn it down, etc. boring little black cape, go sit down.


the better thing is: systems are necessary and monstrous. mercy without structure dies at the first succession dispute. structure without mercy becomes a knife that forgot hands exist.


so the player lives in the bad middle.


the question is not “are you good?”


the question is:


what falsehood are you willing to make official so these people can survive until winter?


that’s the part i’d build first. not the whole grand monk apparatus. not the full village sim. just ten disputes, four ways to encode each ruling, and consequences that come back wearing your own language like a stolen coat.

Monday, 12 January 2026

hypercube editor / parametric scenarios / trinary computer

 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.