[pinterbanget]

Summary

This adds research/p104-fiber-tag-orientation-check/. It independently validates #4154 (unpaidmoon, merged), which the p104-side-correction card lists as still needing validation.

Claim type: an independent finite check with a correction. No new κ.

What it confirms. Every claim that does not depend on orientation:

• the labels are Prove P104 phase-uniform gauge and unavoidable completion tag #3966's;
• no full-address agreement in any of the 10,816 ordered pairs;
• the shared-line split is 5,440 agree / 5,376 disagree;
• the uniform tag a is never a fiber tag;
• the minimum Hamming distance is 6, only at labels 4 and 32;
• the witness Hamming distance is 20.

What it corrects. #4154's code builds its row tag as z ⊗ (column image) and its column tag as (row image) ⊗ z. Its own README, and #3966's PROOF.md §4, define the opposite: row correction z ⊗ q_j with q_j = O p_j, and column correction r_i ⊗ z. #3954's witness lines decode as 386 ⊗ 386 (row port) and 405 ⊗ 386 (donor port), so the witness pair is row 386, column 405.

In that orientation:

At the witness	#4154	Source orientation
Shared-line flips (row, column)	(0, 1)	(1, 0)
Does a match the row tag on the shared line?	no	yes
Separating cell	bit 11	bit 19 = (2, 1); bit 11 does not separate

• The census sentence needs one word changed. "a matches the row tag exactly when the column label is class 3" should say the row label. The count is 4,992 either way.
• The full-address obstruction is unaffected. a is never a row tag.
• The shared-line conclusion is reversed. Record the P104 fiber-tag mismatch on one row/column pair. #4154 says a helper carrying a "does not match this row owner" on the shared line. That conclusion is the premise of its open question 2.

Control. Evaluating #4154's transposed objects reproduces its 0/1 flips, its helper mismatch and its bit 11 exactly, with the census unchanged. The difference is therefore exactly the transpose.

Provenance note. #4154 says #3966's RESULT.json "is not a Git object on this base". It is on main at research/p104-phase-uniform-gauge/source/RESULT.json, with the pinned SHA256.

Evidence

• Standard library only; runs in about 2 seconds.
• Imports nothing from Record the P104 fiber-tag mismatch on one row/column pair. #4154. Tensors are represented as (α, β) cell sets.
• Labels are read from Prove P104 phase-uniform gauge and unavoidable completion tag #3966's file. The orientation comes from Prove P104 phase-uniform gauge and unavoidable completion tag #3966's text and research: derive P104 two-axis endpoint intertwiners and cost gate #3954's witness vectors.

python3.12 -B scripts/research_package.py --package research/p104-fiber-tag-orientation-check --check-only
  PASS_PACKAGE_METADATA_AND_PINS (not a mathematical acceptance check)
python3.12 -B scripts/research_package.py --package research/p104-fiber-tag-orientation-check
  PASS_REGISTERED_RESEARCH_PACKAGE research/p104-fiber-tag-orientation-check outcome=completed

Head: a6e02c006ccbb6612c3a6e8fcae8ae49c2f474c4.

Not covered

• Native routing, owners, dirty return, precision and the 44,096 excess.
• Which orientation a future caller should adopt. This package only matches the sources' definitions.

Task and overlap

• Card: p104-side-correction, refined, priority revision 7.
• Base: canonical main 12fa8c5c844be21bfefb38a19503fbd867a11244.
• No PR or discussion had independently checked Record the P104 fiber-tag mismatch on one row/column pair. #4154.

Credit

The seam comparison is unpaidmoon's (#4154). The labels, the gauge and the corrections are #3966's (ToikY); the witness is #3954's (Clear1nsight). Prepared by Ryanson Jonathan with substantial Claude Code (Claude Opus 5.5) assistance. Apache-2.0.

🤖 Generated with Claude Code