Skip to content

Prove the selected #5599 scalar instructions lift at every column mul… - #5837

Merged
Mister-Meeseeks merged 1 commit into
CrocSwap:mainfrom
unpaidmoon:cursor/native-scalar-chart-lift-ad37
Oct 27, 2026
Merged

Mister-Meeseeks merged 1 commit into
CrocSwap:mainfrom
unpaidmoon:cursor/native-scalar-chart-lift-ad37

Conversation

@unpaidmoon

@unpaidmoon unpaidmoon commented Oct 27, 2026 •

Copy link
Copy Markdown
Contributor

Change and mathematical scope

Scoped proof of one implication of assumption native_all_columns for the selected #5599 word. No new κ. The selected conditional result stays κ = 62009/20000000.

Every SCALE, ADD and COPY in research/p104-helper-subspace-splits/WORD.json.gz acts on logical chart coordinates by the same rational matrix at every column multiplicity f ≥ 1, with the scalar coefficient outside the tensor power. The 342 events added to the pinned PR5565 word are 300 ADDs and 42 SCALEs, so the thirteen helper splits are inside that class. The 3,447 MOVEs remain inside the assumption.

If two charts differ by -I, the logical correction is (-1)^f. Width 2 then matches the width-free replay and widths 1 and 3 do not, so a finite width does not replace the common-chart hypothesis.

Status: proof of one caller-bound implication. This is not a paid-profile change, a κ improvement, or a discharge of the whole assumption.

Remaining obligations: MOVE chart operators with absolute phase; spectators, whole-job adapters and the fixed-tape tariff; owned_erasure, compiler, precision_tape and analytic_recovery.

Task and overlap

Validation

Python 3.12, standard library. From the repository root:

python3 -B research/native-scalar-chart-lift/verify.py
python3 -O -B research/native-scalar-chart-lift/verify.py
python3 -B scripts/research_package.py --package research/native-scalar-chart-lift

Observed marker:

PASS_NATIVE_SCALAR_CHART_LIFT adds=21247 scales=757 copies=8 moves=3447 common_chart_violations=0 split_delta_add=300 split_delta_scale=42 split_delta_move=0
PASS_REGISTERED_RESEARCH_PACKAGE research/native-scalar-chart-lift outcome=partial

The all-f statement is the induction in PROOF.md. Widths 1–3 check the tensor implementation. The checker rejects a cross-chart synthetic ADD, a stale frame after MOVE, and a one-step retag of the first selected ADD.

Not run: make frontier5599-verify, native Clifford replay, Lean, Docker.

Credit

Ali Romman supplied the selected word (PR #5599). Pierre-Loic Doulcet / hexapode supplied the PR #5565 word. The caller lineage is credited in research/p104-helper-subspace-splits/README.md. The selected review is Douglas Colkitt's docs/research/pr-intake-evidence/pr5599-completion/PROOF-REVIEW.md. Mark / Matarener supplied #5442. #4610's admission note records the older scalar-outside-tensor convention.

Prepared by unpaidmoon with substantial Cursor Grok assistance. Apache-2.0.

Open next questions

Can the 3,447 retained MOVE chart changes of this same word be lifted by g ↦ g^{(⊗)f} with absolute phase, as group identities rather than width-1 recipe checks?

That is the remaining algebraic content of native_all_columns for this caller. Census one distinct frame pair through the pinned clifford_exact canonical frames and apply the #5442 lifting lemma, including the action on |0>. The selected native checker compares width-1 signed Pauli recipes only. A passing census would still leave spectators, adapters and the fixed-tape tariff open.

…lumn multiplicity.

The common-chart ADD, SCALE and COPY events of the helper-subspace word,
including the 342 events added to the PR5565 word, act by an f-independent
rational matrix. A chart difference of -I matches the width-free replay at
even widths and misses it at odd widths. MOVE operators remain open.

Prepared by unpaidmoon with substantial Cursor Grok assistance.

Co-authored-by: unpaidmoon <unpaidmoon@users.noreply.github.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants