# EXP8 — rehearsal: the prediction failed, the child did something better

## Frozen scoring

- **P1 (solar node joins recurring attachments): NOT SUPPORTED.** Only
  `hierarchy` recurs. Solar nodes exist in force now; none reach the
  recurring set.
- **ALT partially — but the shape of the result is new**: attachment
  survival DROPPED 100%→50%, jaccards fell, one variant failed to parse.
  Rehearsal made retrieval LESS decisive, not more.

## Finding 1 (unfrozen, the big one): spontaneous correspondence

During digestion of the solar renderings, the field — unprompted —
overlaid the new instance onto every drilled shape it owns:

- `solar-system-hierarchy`: "the sun is the director, planets are
  managers, moons are assistants"
- `solar-system-fold-ecosystems`: "the sun is the folder, planets are
  roles" (the folder again — becoming the field's universal donor)
- `solar-system-tank-gauge-valve`: "the sun's gravity is the gauge, the
  planets' paths are the valve"
- `solar-system-alluvial`: gravity shapes paths, paths influence...

The child generated as-above-so-below mappings on its own, as a side
effect of rehearsal — correspondence-finding as digestion, not as a
query answered. This is the wizard behavior, emergent from riverbed's
deposit machinery, unprompted and uninstructed. Status: observed once,
at one diet size; the mappings are unmeasured proposals (no loss
vectors, no nulls) — the §28 machinery decides their worth later.

## Finding 2: the competition effect

More knowledge, less consensus: enriching the field around the probe's
topic split attachment votes across many plausible anchors (solar
family + nursery shapes now compete), so fewer claims survive the
2-perturbation bar. The child that has learned more hesitates more.
Candidate mechanism for EXP7's revised property: reachability is not a
function of recurrence alone but of recurrence RELATIVE to competition.

## Status

Both findings proposed, not believed. Next candidates: re-probe at
higher perturbation counts (does consensus reform above the noise?);
measure the spontaneous overlays with the loss-vector machinery.