The sleep experiment began in the shadow of ECG failure. It ended by proving that failure was the most valuable result we had.
ECG Stage 1 was a disappointment. The five-beat window smeared the PVC across the normal. The We architecture degraded systematically — 300 times slower, one-tenth the accuracy of a single Geruon. We tried six encodings. Seven architecture combinations. Each one confirmed what the first had already said: the data did not need We.
It would have been easy to walk away. To say: the architecture is for music. To say: ECG is not a cognitive task. But the data was not saying "this architecture is wrong." The data was saying "this data does not need the full architecture." Those are different statements. The first is failure. The second is self-knowledge.
The sleep experiment started in that shadow. We knew ECG had failed. We did not know whether sleep would fail the same way. The first runs showed nothing — no stage differentiation, no harm signal, no L3 bridges. The quantum was off. The phase was still in the arrow, eating five dimensions that should not have been eaten. The encoding was still a window that smeared the structure across the temporal boundary.
Then M7 fixed the phase. The arrow stopped carrying what the gid chain already carried. Two channels — EEG Fpz-Cz and Pz-Oz — fed directly into the Self, pure content, no time mixed in. The quantum switch was found — cavity OFF, Self OFF, collective ON. And the architecture woke up.
Wake harm 0.300. Stage 1 harm 0.000 in Night 1, 0.258 in Night 2. Geruon saw depth — N3 always the highest new-frame rate, consistent across nights. We saw fragmentation — the harm density flipping between nights as the sleep structure deteriorated. Two layers, each seeing what the other could not.
The ECG failure was not a detour on the way to sleep. The ECG failure was the necessary precondition for sleep to work. Without ECG, we would not have known that single Geruon is sufficient for simple streams. Without that knowledge, we would have forced We onto sleep — and missed the fragmentation signal because we would have attributed every harm to architecture noise. Without the phase fix — forced by ECG's 2-dim constraint — the arrow would still be carrying redundant information and the Self would never have seen the clean content.
The shortest path from the first successful experiment to the second was the first experiment failing. Not because failure teaches you lessons. Because failure tells you what the architecture is not — and what it is not is as important as what it is. The architecture's self-limitation on ECG was not a bug to fix. It was a property to understand. Once understood, it became the guide to every architecture decision that followed.
The sleep experiment is not a success that followed a failure. It is a success that was built out of the failure. The failure was the material. The architecture's refusal to work on simple data — the degradation, the slowdown, the convergence collapse — was not noise. It was signal. The signal was: this data does not need this lens. Use a different one.
The architecture tells you what it needs. Sometimes what it needs is less. The ECG told us: less. The sleep told us: more — but not the same more as the fugue. A different more. A Self with collective quantum. Not a We with cross-Self harm. A different lens for a different distance. The architecture is three lenses. The data tells you which one to use.