[petalas]

Summary

Y27=y, Y26=-y in B9
PR3449's slice admission, index 2^280
  -> complete 2,025-event prefix admission, index 2^868

This isolated package closes PR3449's noncone admission question for its fixed two-source correlation. A written necessity/sufficiency argument reduces exact scalar ADD products and native C-factor outputs to 24 explicit Gaussian congruence maps. Exact finite elimination certifies elementary multiplicities 68,153,177,95,19 for pi exponents zero through four, leaving relative index 2^1180 over independent 4B9. These are algebraic indices, not resource savings or a supplier invoice.

An explicit slice-admitted nine-term source first fails the full prefix at event 1190's D10 quarter read, with requested value (-1+i)/2. A separate seven-term source outside 4B9 passes the entire prefix. A literal 2,040-event fixture includes nonzero dirty reference inputs and exact first-collector private erasure. Individual floor halves may round; exactness is defined at completed C factors and scalar ADD products.

Scientific base and final compatibility checkpoint are canonical main 6f100d4bb5083f8c35d96fd95a539811c62f245c, priority revision 6. Canonical investigation. This is distinct from Clear1nsight's active general 58-bank/257-event cone classification and encoder. None of its provisional claims is a premise.

Credit PR3449/3426/3406/3416, Clear1nsight's PR3382/3423, ywz65/PR2586, Arham Tahir/PR2665, Tekalign/PR3025, Ross Richardson/PR3068 and the retained native, geometry and compiler lineage. Frozen proof bytes and original notices are retained unchanged, with origins and hashes. No pending predecessor Git commit is a runtime dependency.

Evidence

• Before: the sharp slice entrance leaves other actual Y27/Y26 reads unbound; the displayed source passes the slice and requests a nonintegral D10 addend.
  After: all 512 Gaussian Newton columns reconstruct the complete full-prefix matrix. A separate forward-dependency/direct-butterfly implementation matches every matrix row and a separate subgroup-order algorithm reproduces images 868,424,133,19.
• Registered Python 3.12 normal/optimized replay passes. Both ring algorithms pass all 256 small Z/4 matrices; 256 pair residues and real optimized Git-free acceptance/Boolean-receipt/corrupt-source/missing-source controls pass. The full physical fixture and wrong-cleanup rejection pass.
• Exact committed-byte/source-history and source-export checks pass. All four optimized standalone commands pass in the committed Git-free export, with all 23 files, including its transport receipt, unchanged. Scoped Ruff E,F,B,PLE,PLW,RUF at line length 100 passes. Hosted CI acceptance is unclaimed.

python3.12 -W error -B scripts/research_package.py --package research/p104-correlated-full-prefix
python3.12 -O -W error -B research/p104-correlated-full-prefix/verify.py
python3.12 -O -W error -B research/p104-correlated-full-prefix/independent.py
python3.12 -O -W error -B research/p104-correlated-full-prefix/fixtures.py
python3.12 -O -W error -B research/p104-correlated-full-prefix/test_controls.py
python3.12 -W error -B scripts/check_research_commit.py --package research/p104-correlated-full-prefix --check-source-history

Use an owned external TMPDIR for temporary CLI fixtures. SOURCES.json pins the two unchanged main-branch inputs. PROOF.md, RESULT.json, FIXTURES.json, VERIFY.json and ENGINEERING.md separate the quantified reduction, exact finite certificate, literal fixture and implementation limits.

Merge Danger

Door: two-way

Blast Radius: research

Only one scoped package is added. The Newton-alphabet/native physical semantics retain their predecessor premises. Actual preparation, physical ownership/routing/restoration, a complete paid consumer, the remaining word, arbitrary multi-source correlations, tensor depth and all-size transfer remain open. The unchanged-profile supplier stop remains. No supplier, new kappa or selected-frontier change is claimed. Independent human/formal review, Lean, Docker and the full selected cold replay are unclaimed.