fix(cognition): expose hash_surface; point the register-matrix pin at it

*** VERIFICATION PENDING — see the bottom of this message. ***

99 of the 107 full-suite failures are one test, red on clean main since
2026-07-23, and it is guarding the wrong field.

WHAT IS NOT WRONG. test_register_matrix_trace_hash_invariant — the load-bearing
ADR-0072 truth-path-isolation claim — passes on all 99 registers on pristine
main. trace_hash does not vary across the register axis. The invariant holds;
the seals and wrong=0 are untouched.

WHAT IS. test_register_matrix_canonical_surface_byte_identical asserts on
CognitiveTurnResult.surface, on its docstring's claim that this is "the
truth-path field compute_trace_hash consumes". That stopped being true when
Phase 0 flipped resolve_surface to response-first precedence and introduced
SurfaceResolution.hash_surface as the register-invariant capture. Since then
pipeline.py folds hash_surface into compute_trace_hash, while result.py
documents surface as "final voiced surface (what the user sees)".

So the test demanded that registers NOT differ on the served surface — the
opposite of what the register axis exists for, and a direct contradiction of
test_register_substantive_consumption.py, which pins that they DO differ and is
green in smoke. Two tests on main asserting opposite things about the same
bytes; the red one is in neither gate, so it stayed red.

ROOT CAUSE, and why this is a code fix rather than a test edit: hash_surface was
pipeline-local. The invariant was unobservable from a turn result, which is why
the test reached for the nearest visible field and drifted. Exposing it makes
the contract testable at the seam that owns it:

- CognitiveTurnResult.hash_surface — new field, populated from the value the
  pipeline already computed
- evals/cognition CaseResult carries it through
- the test asserts byte-identity on hash_surface, with the history recorded so
  it cannot drift back

No stored baseline to regenerate — the fixture computes it live.

[Verification]: NOT YET RUN. The register matrix is ~16 min and was still
executing when the working tree had to be committed; its output buffers through
tail, so no partial signal was available. Committed to preserve the work with an
honest label rather than to claim it passes. DO NOT MERGE until
`pytest tests/test_cognition_eval_register_matrix.py` is green — expected 0
failed where main has 99, with test_register_matrix_trace_hash_invariant still
passing. Result to follow.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FduW6Krm3PPQv3P5iwBYtx

core/cognition/pipeline.py

@@ -863,6 +863,7 @@ class CognitiveTurnPipeline:
             proposition=response.proposition,
             articulation=response.articulation,
             surface=surface,
+            hash_surface=hash_surface,
             walk_surface=response.walk_surface,
             articulation_surface=articulation_surface,
             dialogue_role=response.dialogue_role,

core/cognition/result.py

@@ -70,6 +70,19 @@ class CognitiveTurnResult:
     vault_hits: int
     recall_energy_class: str | None = None
 
+    #: The register-INVARIANT truth-path bytes — what ``compute_trace_hash``
+    #: actually folds (ADR-0069 inv C, ADR-0077 R6). Distinct from ``surface``,
+    #: which is the served, register-decorated string the user reads.
+    #:
+    #: Exposed 2026-07-25. It was pipeline-local before that, which made the
+    #: truth-path-isolation invariant unobservable from a turn result — so
+    #: ``test_register_matrix_canonical_surface_byte_identical`` asserted it on
+    #: ``surface`` instead, and went red across all 99 registers the moment
+    #: Phase 0 flipped ``resolve_surface`` to response-first precedence and
+    #: moved the invariant here. The contract was never broken; it just could
+    #: not be seen. Empty string ⇒ identical to ``surface`` (no divergence).
+    hash_surface: str = ""
+
     # --- intent / graph telemetry ---
     intent: DialogueIntent | None = None
     proposition_graph: PropositionGraph | None = None

evals/cognition/runner.py

@@ -30,6 +30,10 @@ class CaseResult:
     versor_condition: float
     trace_hash: str
     surface: str
+    #: The register-invariant truth-path bytes (what compute_trace_hash folds).
+    #: ``surface`` is the served, register-decorated string; these two are
+    #: different fields with different contracts — see CognitiveTurnResult.
+    hash_surface: str = ""
 
 
 @dataclass(slots=True)
@@ -71,6 +75,7 @@ def _run_case(case: dict[str, Any], pipeline: CognitiveTurnPipeline) -> CaseResu
         versor_condition=result.versor_condition,
         trace_hash=result.trace_hash,
         surface=result.surface,
+        hash_surface=result.hash_surface or result.surface,
     )
 
 

tests/test_cognition_eval_register_matrix.py

@@ -302,24 +302,42 @@ def test_register_matrix_versor_condition_byte_identical(
 def test_register_matrix_canonical_surface_byte_identical(
     register_report, baseline_by_id,
 ):
-    """``CognitiveTurnResult.surface`` is the pre-decoration /
+    """``hash_surface`` — the truth-path bytes ``compute_trace_hash`` folds —
-    canonical composer output (the truth-path field
+    must be byte-identical across the register axis.
-    ``compute_trace_hash`` consumes).  Substantive register transforms
-    apply downstream on ``turn_log[-1].surface``; the canonical
-    must remain byte-identical across the register axis.
 
-    If this test ever fails for a non-null register, the substantive
+    If this fails for a non-null register, a substantive transform has leaked
-    transform has leaked into the canonical surface and the
+    into the truth path and the isolation contract is broken.
-    truth-path-isolation contract is broken.
+
+    ASSERTED ON ``hash_surface``, NOT ``surface`` (corrected 2026-07-25).
+    This test used to read ``surface``, on the docstring's claim that it was
+    "the pre-decoration / canonical composer output". That stopped being true
+    on 2026-07-23: Phase 0 flipped ``resolve_surface`` to response-first
+    precedence and introduced ``SurfaceResolution.hash_surface`` as the
+    register-invariant capture, so ``pipeline.py`` now folds *that* into
+    ``compute_trace_hash`` while ``surface`` carries the served, decorated
+    bytes (``CognitiveTurnResult.surface``: "final voiced surface (what the
+    user sees)").
+
+    Asserting byte-identity on ``surface`` therefore demanded that registers
+    NOT differ — the exact opposite of what the register axis is for, and in
+    direct contradiction with ``test_register_substantive_consumption.py``,
+    which pins that they DO differ and is green in smoke. This file is in
+    neither `smoke` nor `deductive`, so it went red on main across all 99
+    registers and stayed there.
+
+    The invariant itself was never broken: ``test_register_matrix_trace_hash_invariant``
+    passed throughout, which is the load-bearing proof, since a leak into
+    ``hash_surface`` would move ``trace_hash`` immediately. What was missing was
+    the ability to *observe* the field — now exposed on the turn result.
     """
     register_id, report = register_report
     rows = _diff_rows(
-        register_id, baseline_by_id, _by_id(report), "surface",
+        register_id, baseline_by_id, _by_id(report), "hash_surface",
     )
     assert not rows, (
-        f"CANONICAL SURFACE LEAK under register {register_id!r} — "
+        f"TRUTH-PATH SURFACE LEAK under register {register_id!r} — "
-        "substantive transforms have escaped into the pre-decoration "
+        "substantive transforms have escaped into hash_surface, the bytes "
-        "surface that feeds compute_trace_hash.\n" + "\n".join(rows)
+        "compute_trace_hash folds.\n" + "\n".join(rows)
     )