EXP-064NEGATIVE
Canonical GPU engineering / SHA-256d

Pilot screening of header templates

A small pilot can select headers whose disjoint holdout has a higher below-target density than equal-work random allocation.

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
ReproducibilityPreserve disjoint pilot/holdout windows, charge pilot work and repeat both independent panels and controls.
01 / Question & hypothesis

What was tested?

A small pilot can select headers whose disjoint holdout has a higher below-target density than equal-work random allocation.

02 / Scientific basis

Why the test is meaningful

Pilot hashes are real work and must be charged. Transfer requires an unseen holdout and comparison with many equal-budget random selections.

net gain=(pilot+selected holdout hits)/(same total random work)top 8 of 64 by pilot B8transfer requires panel replication
Pilot screening of header templatesVisual reading of the published metrics and gates for EXP-064; it summarizes the registered result, not a mining advantage.EXP-064 / PILOT SCREENING OF HEADER TEMPLATESAGGREGATE B8 NET GAIN1.0010SPEARMAN CORRELATION−0.0955PERMUTATION P0.8517
FIGURE / RESULT READINGVisual reading of the published metrics and gates for EXP-064; it summarizes the registered result, not a mining advantage.
03 / Method

How it was tested

Use two independent panels of 64 headers, 2^15 pilot nonces, disjoint 2^16 holdout nonces, top-eight selection and 32 matched random controls.

04 / Observed result

What happened

Aggregate B8 net gain1.0010
Spearman correlation−0.0955
Permutation p0.8517

Across 12,582,912 exact SHA-256d hashes, panel 0 net B8 gain was 1.0094 but panel 1 reversed to 0.9926. Aggregate B8 net gain was 1.0010, Spearman −0.0955 and permutation p=0.8517.

05 / Validation

Exactness and statistical controls

A secondary B16 gain from only 19 selected hits was retained but not promoted because it lacked replication in registered endpoints.

06 / Interpretation

What the result means

Header-template enrichment did not transfer under the frozen protocol and post-hoc retuning is closed.

Limitations

  • B16 was sparse and secondary.
  • Only the registered pilot/holdout sizes were tested.
  • Header selection does not change hash validity.
07 / Reproduction

Evidence trail

Preserve disjoint pilot/holdout windows, charge pilot work and repeat both independent panels and controls.

Canonical variants

CANONICAL-EXP-140PREREGISTERED-CAMPAIGNSEALED-AUDIT

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