EXP-060NEGATIVE
Canonical GPU engineering / SHA-256d

Target-aware selector on partial second-SHA state

A frozen linear score of partial second-SHA state can select nonces that beat uniform completion under equivalent SHA-round work.

Claim statusNO END-TO-END MINING ADVANTAGE DEMONSTRATED

Diagnostic work, exact algebra and local capabilities are not treated as end-to-end mining advantage.

Date2026-08-14
HardwareNVIDIA GeForce RTX 5080 · exact SHA-256d verifier
Run scopePreregistered campaign with sealed result and audit
ReproducibilityFreeze the selected frontier on discovery, apply it unchanged to unseen headers and compare against an equal-round uniform baseline.
01 / Question & hypothesis

What was tested?

A frozen linear score of partial second-SHA state can select nonces that beat uniform completion under equivalent SHA-round work.

02 / Scientific basis

Why the test is meaningful

Selector cost must include the first SHA, the evaluated second-SHA prefix and completion of selected nonces; uniform controls receive exactly the same round budget.

gain_Bk = hits_selected(Bk)/hits_uniform_equal_work(Bk)frontiers={16,24,32,40,48,56}gate: gain≥1.05 on 4/4 unseen headers
Target-aware selector on partial second-SHA stateVisual reading of the published metrics and gates for EXP-060; it summarizes the registered result, not a mining advantage.EXP-060 / TARGET-AWARE SELECTOR ON PARTIAL SECOND-SHA STATEREPLICATION B12 GAIN0.603381×HEADERS PASSING0 / 4TRACER PROBES128 / 128
FIGURE / RESULT READINGVisual reading of the published metrics and gates for EXP-060; it summarizes the registered result, not a mining advantage.
03 / Method

How it was tested

Discover on four headers, freeze round 16 with one-half selection, then replicate unchanged on four unseen headers over 262,144 nonces each.

04 / Observed result

What happened

Replication B12 gain0.603381×
Headers passing0 / 4
Tracer probes128 / 128

The exact tracer passed 128/128 probes plus a 32/32 smoke test. Replication gains were 0.602669× at B8, 0.603381× at B12 and 0.588235× at B16; B12 lost on 4/4 headers.

05 / Validation

Exactness and statistical controls

Uniform controls were work-equivalent, the selector was frozen before replication and all traced digests matched hashlib.

06 / Interpretation

What the result means

This linear-bit selector family cannot repay work invested before selection and is closed.

Limitations

  • Only registered linear 64-bit scores were tested.
  • Other representations require fresh preregistration.
  • No validation/test split was opened after failure.
07 / Reproduction

Evidence trail

Freeze the selected frontier on discovery, apply it unchanged to unseen headers and compare against an equal-round uniform baseline.

Canonical variants

CANONICAL-EXP-136PREREGISTERED-CAMPAIGNSEALED-AUDIT

Source: internally audited canonical reports. Local filesystem structure, private headers and operational identifiers are excluded from publication.