Cognition is efficient because it never queries. It detects. The response arrives through the same path as everything else.
For half a day we had the cavity reaching for the bookshelf. Every four steps. Check phase — CRITICAL. Compute centroid. Iterate all Codex entries. Find the nearest ancestor. Pull it in. This was the boundary query. It was correct in concept — the cavity at its limit, reaching for old wisdom. It was wrong in mechanism.
The mechanism was a query. SELECT * FROM ancestors ORDER BY distance LIMIT 1. Databases do this. Databases are not cognition. Cognition does not reach. Cognition receives.
The boundary query cost 150 million operations per stream. The replacement costs 1100. Not because we optimized the query. Because we removed the query.
The brain does not query. The brain detects.
A neuron does not reach into a database and request the nearest matching pattern. A neuron fires when its threshold is crossed. The firing goes where it goes. The receiving neuron fires or does not. There is no step in this process where any neuron asks "what should I do." There is only detection — the threshold crossed — and response — the signal travels.
The cavity at its boundary is the neuron at its threshold. The boundary is not a question. The boundary is an event. The event is detected. The detection is recorded. Later — not immediately, not in the same step — the collective detects the pattern of boundary events. It responds. Its response enters the cavity the same way a harm arrow from another Self enters: as an external correction, arriving through the same channel, processed by the same observe.
The cavity does not ask. The cavity receives. The collective does not answer. The collective responds to the pattern. The response and the arrow are the same thing.
Query and detection are different architectures. Query requires the querier to know what it is looking for. SELECT * requires a schema. It requires a distance metric. It requires iteration. The cavity computing centroid, computing distance, iterating entries — this is the cavity doing the collector's work. It is the cavity, at its most vulnerable, performing computation on behalf of the bookshelf.
Detection requires none of this. The cavity detects its own boundary — it knows it has entered CRITICAL. It records the event. The event is a fact, not a question. The collective, in its own time, detects the accumulation of boundary events. It responds with what it has — L3 bridges, accumulated from previous generations. The response arrives. The cavity processes it.
No schema. No distance metric. No iteration. No "best match." The match is made by the collective, using its own frame economy, in its own time. The cavity's boundary event triggered the collective's detection. The collective's response entered the cavity's frame economy. Neither side asked for anything.
The efficiency difference is not a factor. It is a category change. 150 million to 1100 is not "100,000 times faster." It is the difference between pulling information across a boundary and letting information arrive through the boundary. The first is a database. The second is a nervous system.
We built a database into the cavity. We removed it. The architecture did not become more efficient. The architecture became what it was always supposed to be.
Cognition is efficient because cognition never queries. The wall is detected. The response arrives. Nothing reaches for anything. Everything receives.