# 004 shuffled-dae — RESULT (harvested 2026-08-01, tier T3-exploratory)

**Verdict: premise held, both predictions about z's order code came out OPPOSITE-of-naive —
forcing order into z made it LESS linearly readable, and no role code emerged.**

## Numbers
- Manipulation check PASSED: reorder tau_decoded **0.447** vs shuffled-input floor 0.028
  (gate floor+0.2 cleared); decode-F1-from-shuffled 0.730; val_exact 0.109 (baseline ~0.03);
  val_f1 0.7329 — bag-F1 EASIER under shuffle, exactly as the manager's contra-prediction
  reasoned (bag metric, no deletions).
- Battery: **surface cross 0.615 [0.597, 0.635] — BELOW baseline 0.706**, opposite of the
  forced-order prediction (>0.85). Primary role cell 0.504 [0.494, 0.515] = chance (as
  predicted). Within-ceiling 0.586/0.687 (also depressed vs baseline 0.664/0.767).

## Interpretation (scoped)
z demonstrably carries enough order information to reorder a full bag (tau 0.45), yet the
battery's linear surface-position probe reads order WORSE than baseline. So the
reconstruction-driven order code is stored in a format the battery's canonical readout
does not linearize (plausibly word-identity↔position associations rather than a global
"who's first" direction), and it DILUTED the linearly-readable surface signal rather than
amplifying it. Data-pressure lesson mirroring 001-003: forcing information INTO z says
nothing about the FORMAT it takes. No role code (4th flat primary).

## Brier: (a) 0.45→F, (b) 0.12→F, (c) 0.55→F (tau 0.447 < 0.6), (d) 0.55→T. Mean **0.180**.

## Caveats / follow-ups
- Tau conflates z-order with decoder LM prior (pre-flagged); the surface-cell drop is the
  cleaner z-side readout and is the notable number here.
- Suggests a probe-format experiment (nonlinear/pairwise readout of order from shuffle-z)
  if promoted; connects to 097 (pooling symmetrization theory: what formats survive).