This website is meant to be read and understood quickly by humans, but is only fully parsable, on a technical level, with the aid of an AI system. Read why →
Loop MMT
Knowledge Architecture · P-real-29

What A Dog Needs To Know

AI Animal → Dog → Border Collie

A KP request that discovers a taxonomy as it's typed:

The Prompt

Ed! I have a cool job for us- I want to create a new KP for 'what a dog needs to know'. I would very much like to have a border collie running around but I would like it to be driven by science, math, and reality as much as Leroy is. I think we need something special here for dogs- for the most part, a cat is a cat is a cat, even when you scale up to bigger ones like tigers and lions. But dogs are different- my border collie Sophie is a VERY different creature from a golden retriever, and from a Great Dane, and from a teacup poodle, so I think we need to have, for dogs, TWO KPs that compose on each other- one for the basic Dog and one for the Border Collie. In fact, now that I've typed that, could there be some kind of Context Mesh here? Are there anythings that cats and dogs SHARE? I don't want to make this system overly complicated just for kicks, so chisel this idea as you work it, but could there be a system like AI ANIMAL->DOG->BORDER COLLIE, and AI ANIMAL-> CAT, just as you could have AI ANIMAL->GIRAFFE? Please KX hard on this- make sure you get all the right ones installed, and then just work the idea over with a Super RCR started off with a ~discovery sensorium and a molly and adderall steep, plus some good tunes on the Liner Notes. Let's make this beautiful with a rock solid foundation. Also, feel free to have space to rework the Cat KP in the end- I am guessing we might come up with better ideas for how to build this in doing the Dog KP, so we'll, of course, roll those improvements back into the Cat KP afterward, so maybe keep track of that stuff along the way. Come out of the Super RCR with the self reviewed plan for how to actually do all the things that you think need to be done to make this happen. Then we will see where we stand there. AX. Let's have Leroy chair this one with Wes helping him out, since, you know, he's a cat.

— Shea Gunther · operator drop, 10.2006

This prompt A Knowledge Pack for what a dog needs to know — grounded in science and reality like the cat. But a dog isn't a cat: breeds differ enough to need two composing layers (base Dog + specific Border Collie), and maybe a shared substrate for what any animal needs.
The aim The AI ANIMAL context mesh — a shared animal substrate with Dog and Border Collie KPs composing on it, and the Cat KP reworked to match.
A dog isn't a catTwo composing layersAI ANIMAL → DOG → BORDER COLLIECat KP reworked

A stated aim that shipped — the AI ANIMAL mesh: a base substrate (what any animal needs) with a Dog KP and a Border Collie KP layered on it, alongside the Cat branch, exactly the AI ANIMAL → DOG → BORDER COLLIE / AI ANIMAL → CAT shape the prompt sketches. The Cat KP was reworked with the improvements the dog work surfaced.

The Moment

The operator wants a border collie in the room, grounded the way Leroy the cat is — and notices, mid-sentence, that a dog is not a cat: “a cat is a cat is a cat” even scaled up to tigers, but a border collie, a golden retriever, a Great Dane, and a teacup poodle are genuinely different creatures. So dogs need two Knowledge Packs that compose — a base Dog and a specific Border Collie.

Then the real question arrives in the act of typing it: is there a Context Mesh here? What do cats and dogs share? He sketches the shape himself — AI ANIMAL → DOG → BORDER COLLIE alongside AI ANIMAL → CAT, the way you could also have AI ANIMAL → GIRAFFE — and asks that the idea be chiseled so it doesn't get complicated for its own sake, with room to roll improvements back into the Cat KP afterward.

The wider frameThe prompt that grew the AI ANIMAL mesh. The operator wants a border collie in the room — grounded in science and reality the way Leroy the cat is — and notices in the act of typing that a dog is not a cat: a border collie, a golden retriever, a Great Dane, and a teacup poodle are genuinely different creatures, so dogs need two composing layers (a base Dog and a specific Border Collie) where a cat needs one. That observation opens a small taxonomy: AI ANIMAL → DOG → BORDER COLLIE alongside AI ANIMAL → CAT, with a shared substrate for what every animal needs. All of it shipped, and the Cat KP was reworked with the improvements the dog work surfaced.

The Anatomy

“a cat is a cat is a cat … But dogs are different” — The insight that forces the architecture: cats collapse to one layer, dogs don't. Breed variance is the reason a single Dog KP won't do.

“could there be a system like AI ANIMAL->DOG->BORDER COLLIE, and AI ANIMAL-> CAT” — The taxonomy, drawn by the operator in real time — a shared substrate with species and breed layers composing on top. This is the mesh that shipped.

“chisel this idea as you work it … I don't want to make this system overly complicated just for kicks” — The constraint on the build: find the simplest structure that captures the real difference, not the most elaborate one.

Computational Profile
Words in the prompt≈ 371
ProvenanceAim — shipped as the AI ANIMAL mesh
The insightA dog isn't a cat — breeds need layers
The shapeAI ANIMAL → DOG → BORDER COLLIE
CategoryKnowledge Architecture
Reading the Operator's Shorthand

The prompt is written in the working vocabulary of the methodology. Here is what the shorthand means:

KP
Knowledge Pack — a self-contained reference document the board can pull in to ground a discussion in a specific domain.
Context Mesh
A composition move where two or more reference documents share a common substrate and layer on top of it — here, a base 'animal' layer that both a dog and a cat build on.
KX
A provisioning reflex: pull in the specific Knowledge Packs a task needs before starting, so the work rests on the right domain grounding.
Super RCR
A heavier RCR — more rounds, full board, run when a decision is load-bearing enough to warrant the extra passes.
Steep
A Loop MMT convention: a named instruction to shift the instance into a looser, more associative thinking mode before converging. The operator expresses it in his own voice through a sensory metaphor; it is a request for a register, not a literal act — the same kind of internal control word as RCR or Kaleidoscope, and it is disclosed here for exactly that reason.
Liner Notes
The methodology's enriched-music layer for the sensorium's audio channel — a named soundtrack that sets the working mood.
sensorium
The methodology's cognitive-environment layer: a composed set of sensory channels (a “~discovery+grove” mix, here) that sets the working mood before the thinking starts.
Wes
A board member whose lens is chaos and complexity science; asking for Wes means run the session in a playful, exploratory register.
P-real-29 · Knowledge Architecture