# Cognition as reducibility-pocket discovery

Wolfram's "What If We Had Bigger Brains?" essay carries a structural claim worth pulling out as its own organizing primitive: cognition operates by finding islands of computational reducibility inside an irreducible system, and intelligence scales with how many islands can be held at once.

The claim decomposes into three pieces:

1. **The world is computationally irreducible at the underlying level.** Wolfram's Principle of Computational Equivalence places the threshold for irreducibility low; most non-trivial systems are irreducible in the strong sense.

2. **Irreducibility is non-uniform.** Within any irreducible system, there are infinitely many pockets of local reducibility — patches where behavior can be predicted without simulating every step. The progress of science, and concept-formation generally, is the discovery of more pockets.

3. **Brain size sets simultaneous-pocket-holding capacity.** A 100-million-neuron brain (cat) holds enough pockets for navigation but not compositional language. A 100-billion-neuron brain (human) holds the rough 30,000 pockets that natural-language vocabularies inventory. A 100-trillion-neuron brain, or a neural net of comparable scale, would hold orders of magnitude more, and the qualitative capabilities that emerge at that scale are open questions.

The claim subordinates cleanly to compression-theory-of-understanding: each pocket of reducibility is a compression target; concepts are the labels for those targets; understanding is the act of holding the target as a compressed handle on a complex domain. The contribution is the scale-claim. Cognitive capability is a function of how many pockets are simultaneously available.

## Why this is its own structural primitive

The compression-theory-of-understanding canonical names the act: understanding as compression. This finding names the world-feature that makes compression possible: pockets of reducibility within an irreducible whole. They are observations at different layers.

- compression-theory-of-understanding answers "what is happening when I understand?"
- pockets-of-reducibility-as-cognition answers "what makes understanding possible at all?"

Both are needed. The first is the methodology; the second is the world-feature that lets the methodology fire.

## What this organizes

Once the pocket-of-reducibility frame is named, several existing nodes read as instances:

- **vocabulary-over-syntax** — vocabulary is the catalog of pockets a community has labeled; syntax is the combinator that lets pockets compose. "Vocabulary beats syntax" is downstream of "more pockets means more cognitive purchase."
- **compression-theory-of-understanding** — as above; this is the layer-pair.
- **purpose-selects-mechanism-from-irreducibility** (proposed canonical) — purpose selects *which* pockets get held; the pocket-density story is the world-level version.
- **language-as-mind-to-mind transport** — Wolfram notes that the function of language is to transport pocket-handles between minds; mismatch in pocket-inventories is what makes communication imperfect.
- **the LLM emergent-concept finding** — when neural nets cluster activations into emergent concepts, they are discovering pockets humans have not yet labeled. The model can carry millions; the language carries thousands.

## Where it breaks

**Falsifier 1: cognition without pocket-discovery.** If a system can navigate the world by brute-force simulation within its operating horizon, the pocket framework does not fire. Most embodied cognition does pocket-find; some narrow control loops may be brute-force.

**Falsifier 2: pocket-counting may not be the right measure.** The claim that intelligence scales with simultaneous-pocket-holding is a hypothesis about what brain size buys. It could buy deeper single-pocket processing (narrower but deeper compression) rather than wider pocket-inventory. The data discriminating the two is sparse.

**Boundary: the world must actually be irreducible.** If the underlying system is reducible, pocket-discovery is just ordinary decomposition, and the framework collapses to standard scientific method. The empirical bite depends on the Principle of Computational Equivalence holding.

## Standing in the graph

This node is subordinate to **compression-theory-of-understanding** at the methodology layer and to **computational-realism-as-substrate** at the metaphysics layer. It sits adjacent to **purpose-selects-mechanism-from-irreducibility** (proposed Wolfram-derived canonical from W5) — the two describe the same phenomenon from different angles: pocket-density (what cognition does) and purpose-selection (which pockets get used).