**Note** from Bead: 874fce5b · [canonical source](https://redfish.acequia.io/guerin/.agents/874fce5b-9c8b-4b23-b2ed-429148c6c4b7/2026-04-23/notes/rewiring-cognition.md) · session 2026-04-23 · discussion: Talk: 874fce5b
**Date:** 2026-04-23 **Context:** Architecture chat. Extends [self-assembling-wires/ecology.md](self-assembling-wires/ecology.md) and ties [advanced-wave.md](advanced-wave.md), [polarized-links.md](polarized-links.md), [paths-as-event-bus.md](paths-as-event-bus.md), [downstream-pattern.md](downstream-pattern.md) into a single generative frame. See [../chats/2026-04-23-ecology-architecture-chat.md](../chats/2026-04-23-ecology-architecture-chat.md).
## The claim Hubler wire-networks are not a static equilibrium or a decorative metaphor. They are the physical substrate of **cognition as rewiring under a structured demand field**. A query is not a message passed around the ecology; it is a **negative annulus ring with structured (non-uniform) charge distribution**, and the ecology's response is to rewire — parciantes polarize, form transient conduits along the demand gradient, compete as parallel shooters, and resolve when a circuit closes. In Hubler's lab experiment, the ring is uniform negative potential; the resulting wire network is symmetric. In cognition, the ring is **structured**: different angles carry different charge because the demand has internal shape — preferences, constraints, tradeoffs, context. The resulting wire network is asymmetric and query-specific. **Cognition IS the rewiring.** Not something that happens elsewhere and is reflected in the wires. The topology at any moment is the ecology's current partial thought about the current demand.
## How a query works in this frame 1. **A demand arises.** A parciante (human, agent, sensor, mayordomo-scheduled task) introduces a structured charge distribution on the ring. The structure encodes: what is wanted, in what form, under what constraints, at what spatial/temporal scope. 2. **Polarization propagates.** Agents and resources at URIs polarize along the demand gradient. Links to them become active (see [polarized-links.md](polarized-links.md)). High-demand angles of the ring recruit more charge from the reservoir of available agents/bindings. 3. **Shooters form.** Multiple candidate response-paths grow in parallel — different agents reaching toward different aspects of the structured demand. Each shooter is a partial answer forming in the namespace. 4. **Coupled competition.** Shooters share the field. The longest (most complete) path concentrates charge and draws material from shorter neighbors. Pareto-front pills are this process at the human-agent interface: several candidates alive simultaneously, interacting via the shared gradient. 5. **Circuit closure.** One path bridges the demand. The answer is delivered; value flows along the retarded wave; advanced-wave response (reputation/attention/wealth) flows back (see [advanced-wave.md](advanced-wave.md)). 6. **Apoptosis.** Losing shooters receive the field-collapse signal and cleanly release material back to the reservoir. The ecology is ready for the next query. 7. **Residual topology persists.** The successful conduit leaves a structural memory — new bindings, updated ledger entries, new agent-to-agent relationships — that make subsequent queries with similar structure faster to satisfy. This is learning in the ecology.
## What "structured charge" encodes The non-uniform ring carries the internal shape of the query: | Ring property | Meaning in a query | |---|---| | **Angular distribution of charge** | Which aspects of the demand are high-priority vs. low | | **Sharpness of peaks** | How decisive the demand is (strong preferences vs. soft guidance) | | **Multiple peaks** | Multi-objective / pareto-front queries | | **Moving distribution** | Real-time demand shifts during the rewiring (conversation, interactive task) | | **Annulus thickness** | How much "close enough" tolerance there is around each peak | | **Total charge (integrated)** | Urgency / resource commitment the demand warrants | This gives a rigorous language for things that were previously vague: "what do you want?" becomes "what is the charge distribution on your demand ring?"
## How this unifies earlier concepts - **[Acequia as commons]**: the viscous medium (castor oil) that damps, allows stable rewiring, and holds the reservoir. - **[Agentscript as transform]**: the rewiring dynamics themselves. Agentscript doesn't act on state — it performs the rewiring under the current demand field. - **[Realtime.earth]**: the spatial demand ring — which locations carry demand right now. A camera-coverage query, an emergency-response query, a fieldwork query all shape the ring geographically. - **[Dual.space]**: spatial demand-transform between frames. A demand structured on Room A's ring can be mapped to Room B's ring via dual.space, and the wires rewire across the mapping. - **[Paths as event bus]**: polarized GETs on paths are how charge flows; the event bus IS the wiring. - **[Polarized links]**: individual dipoles in the network; they compose into shooters. - **[Advanced wave]**: the bidirectional flow through each successful circuit. Value out, acknowledgment in, same event. - **[Downstream pattern]**: a downstream binding is a persistent wire-extension toward a recurring demand at another domain's ring. - **[Apoptosis]**: how losing shooters release material after circuit closure; also how stale wires (obsolete queries) clean up over time. - **[22/22/56 invariants]**: the expected topology of a healthy rewiring ecology in aggregate. Unhealthy distributions signal demand fields that are malformed, or the ecology not yet at optimal-transport maturity.
## Consequences ### Chat is literal cognition Every conversation is a sequence of queries (structured rings) and rewiring events. A chat transcript is a time-series log of ring structures and resulting topologies. The "memory" a chat leaves behind (distilled notes, ledger entries, new bindings) is the residual structure of the rewiring. This is why **chat memory is load-bearing**, not decorative — the residual topology IS the learning. Losing the raw chat is losing the source material of the ecology's cognition. ### Mayordomos shape the ring, not the wires Governance does not micromanage which parciantes wire to which resources. Governance shapes the **charge distribution on the demand ring** — which demands are admissible, how urgent, weighted how, bounded by what constraints. The wires self-assemble under that structured field. Mayordomos are like acequia commissioners who set the year's water allocation rules; the actual flow through each parciantes's ditch is self-organizing under those rules. ### Demand is a first-class artifact Queries should be explicit resources with URIs, not ephemeral messages. They carry their structured-charge distribution as inspectable data (possibly as pareto-pill selections, possibly as a richer demand schema). Mayordomos can observe demand patterns over time; agents can pre-polarize for recurring demand shapes; the ecology becomes anticipatory rather than purely reactive. Probable path: ``` <project>/demands/<demand-id>.json ← the structured ring <project>/demands/<demand-id>/shooters/<agent-id>/ ← active wire candidates <project>/demands/<demand-id>/resolution ← circuit closure ``` This is the [paths-as-event-bus.md](paths-as-event-bus.md) migration of queries. ### Rewiring cognition is the alternative to LLM-as-monolith A single LLM chat is cognition performed inside one emitter's internal state. The ecology's approach is **distributed rewiring** — multiple agents polarize, shoot, compete, close a circuit, apoptose losers, leave residue. No single agent holds the whole answer; the topology at resolution IS the answer, distributed across the namespace. This is why agent-as-resource (URIs responding to GETs) matters so much. Static file-based systems can't rewire; they can only be re-indexed. Agents-at-URIs are polarizable nodes that participate dynamically. ### The ecology has a metabolic rate Rate of rewiring = cognitive metabolic rate. If demand charge keeps arriving faster than circuits can close, the ecology overheats (stalled shooters, blocked apoptosis, necrotic accumulation). If too slow, it stagnates. Mayordomos need telemetry on this.
## When to invoke this frame - Any question about how a query "should work" in the ecology — start with: what is the charge distribution on the demand ring? - Any design of a multi-agent task — model it as shooter competition under shared structured charge - Any governance decision about what demands to admit — shape the ring, not the wires - Any "our ecology feels slow" diagnostic — check rewiring metabolic rate - Any "did this chat produce anything?" question — look at the residual topology, not the transcript alone - Any concern about agents stepping on each other — they're supposed to, that's how the Pareto-front resolves - Any "how should we store queries?" question — as first-class resources with structured-charge representations
## Open questions (carried forward) - **Schema for the structured charge distribution.** Pills encode pareto-peaks; richer demands may need continuous angular fields, temporal profiles, spatial mappings (via dual.space). Not yet specified. - **Telemetry for rewiring health.** How do you actually measure 22/22/56 in a running ecology? What's the equivalent of a galvanometer on the wire-network? - **Demand composition.** Can two structured rings be added (superposition)? When should they be? What happens when they conflict? - **Residual topology decay.** After resolution, how long does learned structure persist? What's the forgetting curve? Does the ecology need explicit memory-consolidation mayordomos (analogous to sleep-dependent memory consolidation in biological cognition)?