Skip to content

Preserve native-history and odd-E8 research handoff (no new kappa) #5581

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
Mister-Meeseeks merged 1 commit into from
Oct 25, 2026
Merged

Preserve native-history and odd-E8 research handoff (no new kappa) #5581

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
58 changes: 58 additions & 0 deletions docs/research/handoffs/ali-romman-20261025/EXPERIMENTS.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,58 @@
# Stopped families and evidence limits

This table is a continuation guide. Each negative result applies only to
its stated model. The full local record is retained in the project archive.
Names in the last column refer to files in that record. They are not new
public verification claims.

| Family | Result and stop rule | Evidence |
| --- | --- | --- |
| Original clean-word snapshot caches | Equal-frame, monomial, and bounded operator-unit reuse gave no paid saving. Do not repeat the same cache class. | [Finding #5355](https://github.com/CrocSwap/integer-mult-bounds/discussions/5355); `snapshot_discovery.py`, `span_discovery.py`, `all_anchor_discovery.py`, `module_discovery.py`, `unit_module_discovery.py`. |
| Changed permutation codes and triple DAGs | The declared complete optimistic bills lose. This is not a theorem about all codes or DAGs. | [Finding #5381](https://github.com/CrocSwap/integer-mult-bounds/discussions/5381). |
| Private-helper ownership cuts | The single-helper and group-cut tests gave no profitable candidate. Later owned-subset tests repeat this family. | [Released finding](https://github.com/CrocSwap/integer-mult-bounds/discussions/5125#discussioncomment-19154948); `ownership_cut_screen.py`, `owned_subset_ssa.py`. |
| Native scalar and map cancellation | A useful circuit reduction is in PR #5403. Retiming the paid paths gave no paid gain. | [Package](https://github.com/CrocSwap/integer-mult-bounds/pull/5403); [retiming result](https://github.com/CrocSwap/integer-mult-bounds/discussions/5125#discussioncomment-19155692). |
| Repeated two- and three-pair controllers | The 266,584-case bounded test and exact cost checks gave no improvement. | [Finding](https://github.com/CrocSwap/integer-mult-bounds/discussions/4611#discussioncomment-19156205); `repeated_stage_screen.py`, `repeated_stage_verify.py`. |
| Coupled native inverse | Both inverse compositions are constructed without extra Gaussian calls. Ordinary inverse-child costs remain charged. | [PR #5411](https://github.com/CrocSwap/integer-mult-bounds/pull/5411). |
| Small coupled bridge grammars | Shared banks, full bridges, three observations, uncoded middle calls, and mode transport gave no complete funded caller. | [Released caller record](https://github.com/CrocSwap/integer-mult-bounds/discussions/4930#discussioncomment-19156336); `coupled_*bridge.py`, `coupled_*transport.py`, `coupled_four_bank.py`, `coupled_three_observations.py`. |
| One completed-Q catalyst | The declared 188-state binary graph retains the 5/3 bound. This does not cover more general callers. | `retained_q_catalyst.py/json`. |
| Static native-history completion | Scalar and finite-field chart spans gave no funded repair. Cheaper floating candidates failed exact consistency. | `native_history_completion.py`, `native_history_joint_rank.py`, `native_history_repair_*`. |
| Later native continuation | Tests at 2,469 cuts, including one- and two-readout choices, gave no funded correction. Scalar specialization is a strict scope limit. | `native_one_query.py`, `native_two_readouts.py`, `native_temporal_injection.py`, `native_temporal_query_tree.py`. |
| Fixed-source guard cut | Rank-below-nine additive insertions cannot change the eight terminal modes while all original zero guards stay zero. | `NATIVE-GUARD-CUT.md`, `native_guard_cut.py/json`; [public scope](https://github.com/CrocSwap/integer-mult-bounds/discussions/4930#discussioncomment-19156336). |
| Conditional query ports | Eight paid H1 jobs can expose eight H1 responses. The isolated bill is neutral. A complete caller is absent. | `native_guard_query_ports.py/json`. |
| Fixed-source full-chart observations | Eight exact zero-address witnesses make all 5,897 allowed observations vanish but retain the desired modes. Lower charts and changed sources are outside the statement. | `native_full_chart_kernel.py/json`, `native_zero_anchor.py/json`. |
| Mode observer and encoded query variants | The specified five-call, changed-encoder and feedback tests did not produce a caller. They do not exclude arbitrary observation schemes. | `caller_mode_observability.py`, `caller_changed_query_encoder.py`, `caller_query_feedback.py`, `five_word_handoff.py`. |
| Projection swaps and mode exchanges | 8,640 projection-swap cases and 3,840 rational mode-exchange cases gave no endpoint. | `projection_swap_caller.py/json`, `general_mode_exchange.py/json`. |
| Three-call code transport | Two 27,648-schedule tests, one with 55,296 target-parity cases, gave no endpoint. Stop this grammar. | `code_transport_caller_q4.json`, `code_transport_queries.py/json`. |
| E8 plane probes | A candidate Gram product has forbidden nonzero entries. The shifted alternatives do not give a useful tariff. | `e8_plane_screen.py`, `e8_plane_products.py`, `e8_alternating_plane_screen.py`. |
| Direct-sum product planes | Two exact factors give tariffs 15,232 and 15,552. Neither is claimed optimal. Rank does not imply a tariff floor. | [Finding](https://github.com/CrocSwap/integer-mult-bounds/discussions/1587#discussioncomment-19161781); [separate reproduction](https://github.com/CrocSwap/integer-mult-bounds/discussions/1587#discussioncomment-19168830). |
| Odd E8 direct sums | The decoder and its declining tariff are constructive. Separated complete-helper lifts fail the declared cost gate. Joint producers are outside that obstruction. | [PR #5439](https://github.com/CrocSwap/integer-mult-bounds/pull/5439). |
| Synchronized tensor callers | The favorable physical-wire cost screen still loses. It is a floating planning test, not an all-schedule theorem. The unbound earlier receipt is superseded. | `tensor_fusion_screen.py/json`; `tensor-fusion-screen-before-source-binding.json` is historical only. |
| Fixed P112 query support | The inherited parity-projector proof gives block rank exactly 9r, also for the stated eight-port puncture. Changing block size does not escape that support constraint. | [Attributed finding and certificate](https://github.com/CrocSwap/integer-mult-bounds/discussions/1587#discussioncomment-19167039). |
| F4 Hermitian labels | Dimensions 7–9 exhaust in the author's finite search; a dimension-10 positive control passes. No independent UNSAT certificate was produced. | `f4-gate/ROUND.json`, `f4-gate/search_hermitian.py`; source geometry, field skeleton and certificate code remain credited inputs. |
| Ternary labels | Several bounded searches exhaust; one dimension-nine search times out. No new ternary positive control or independent UNSAT certificate was run. | `f4-gate/search_ternary.py`, `f4-gate/search_ternary_asymmetric.py`. |
| Coxeter–Todd geometry | The direct masks lose their copy-mass gate. Dimension-15 and dimension-16 fitting searches time out. A timeout gives no impossibility result. | `coxeter_todd_gate.py/json`, `coxeter_binary_fit.py`, `geometry_ideal_gate.json`. |
| Unchanged p19 invoice | The native letter and caller letter differ. The omitted 4,256 H1 jobs make the repaired bill lose. Existing reviews already record this defect. | [PR #5440](https://github.com/CrocSwap/integer-mult-bounds/pull/5440); no duplicate audit claim. |
| Small E retained-history fixture | Exact rational spans miss the two negative targets after all retained returns and all 76 one-axis queries. The spectral control passes. The actual large caller is outside this model. | [Preserved code and receipt](e-history/README.md). |

## File and status decisions

The local archive keeps superseded receipts as history. In particular, do
not use the earlier tensor bill before source binding, the rejected floating
repair candidates, or timed-out geometry searches as positive evidence.
The native-gauge and odd-E8 package drafts are superseded by their merged
PRs. Export-check scripts and publication drafts are verification records,
not new mathematical contributions.

The primitive rank proof in the P112 support finding is inherited. The
new check binds its certificate to the literal source and the stated
puncture. Do not claim a new general parity-projector method.

Border-product, spectral-renormalization, lattice-extension, bit-only
outer, and other unbuilt ideas were considered but not implemented.
They are not completed experiments. A future task must state a changed
premise and a complete cost test before expanding one of these routes.

All scopes are released. No publication or packaging task is left open by
this record. The complete negative E decoder, caller pricing, and all-size
admission remain research obligations. Their absence is not evidence of a
universal obstruction.
139 changes: 139 additions & 0 deletions docs/research/handoffs/ali-romman-20261025/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,139 @@
# October 23–25 contribution handoff

This record closes the publication work for two research sequences and the
later native-history and odd-E8 investigation. It preserves the result
links, stop decisions, and the last small retained-history test.

No complete kappa increase was obtained in this investigation. At canonical
main `5e0052ca0aa3036a3afa8b0438927377c8d97d53`, the
[selected record](https://github.com/CrocSwap/integer-mult-bounds/blob/5e0052ca0aa3036a3afa8b0438927377c8d97d53/certificates/selected-result.json)
is `605457647/200000000000 = 0.003027288235`. Its five framework assumptions
remain. This handoff does not change that record or claim a new supplier.

## Delivered results

The [first sequence](https://github.com/CrocSwap/integer-mult-bounds/discussions/1166#discussioncomment-19148608)
contains ten results: one PR and nine self-contained Discussion proofs and
checks. The
[second sequence](https://github.com/CrocSwap/integer-mult-bounds/discussions/1166#discussioncomment-19151557)
contains ten results, six of which have focused PRs. Those indexes keep the
scope, proof, source credit, and verification limits of each result. A later
publication of the same result is not counted as another result.

Three later PRs supply native-map cancellation, a same-paid native inverse,
and an odd-E8 decoder with a scoped lift obstruction. The following ten PRs
were merged by maintainers before this handoff. All ten final-head hosted
runs report success. No unresolved inline review thread was present at the
closeout check.

| PR | Result | Exact package | Hosted check |
| --- | --- | --- | --- |
| [#5255](https://github.com/CrocSwap/integer-mult-bounds/pull/5255) | Sharp graph-support inverse norm | [Source](https://github.com/aliromman/integer-mult-bounds/tree/d63d8848442acc0bdc7b62578e12726c530af3fa/research/sharp-banded-resolvent) | [Success](https://github.com/CrocSwap/integer-mult-bounds/actions/runs/39968325269) |
| [#5321](https://github.com/CrocSwap/integer-mult-bounds/pull/5321) | Lower local inverse-code tape bound | [Source](https://github.com/aliromman/integer-mult-bounds/tree/85b70f45078dfa9dc1f865b0fd0c4e8055ddd540/research/factored-code-tape) | [Success](https://github.com/CrocSwap/integer-mult-bounds/actions/runs/39993985934) |
| [#5322](https://github.com/CrocSwap/integer-mult-bounds/pull/5322) | Inverse-pair error cancellation | [Source](https://github.com/aliromman/integer-mult-bounds/tree/38065821a76108dc512391b124ffb7b64feed93b/research/physical-pair-error-transfer) | [Success](https://github.com/CrocSwap/integer-mult-bounds/actions/runs/39998296588) |
| [#5323](https://github.com/CrocSwap/integer-mult-bounds/pull/5323) | Exact rounded-code recovery state | [Source](https://github.com/aliromman/integer-mult-bounds/tree/f9c722011c61779b1b79e9e38c0aa749b933000d/research/inverse-code-recovery-tags) | [Success](https://github.com/CrocSwap/integer-mult-bounds/actions/runs/39998305100) |
| [#5325](https://github.com/CrocSwap/integer-mult-bounds/pull/5325) | Restored Tensor error join | [Source](https://github.com/aliromman/integer-mult-bounds/tree/bea42af24ee9850949a4b48d329a7dc85de72b11/research/tensor-restored-error-join) | [Success](https://github.com/CrocSwap/integer-mult-bounds/actions/runs/39994953330) |
| [#5333](https://github.com/CrocSwap/integer-mult-bounds/pull/5333) | Full local residual budget | [Source](https://github.com/aliromman/integer-mult-bounds/tree/720b7ecc76b605e4a3f9b90826f2544c6ba9e59e/research/physical-pair-full-residuals) | [Success](https://github.com/CrocSwap/integer-mult-bounds/actions/runs/39996148971) |
| [#5341](https://github.com/CrocSwap/integer-mult-bounds/pull/5341) | Single crossing for fixed p16 partitions | [Source](https://github.com/aliromman/integer-mult-bounds/tree/b0edd31a04fe4705846df7673b4a0ec2442260d0/research/fragmentation-single-crossing) | [Success](https://github.com/CrocSwap/integer-mult-bounds/actions/runs/39997680141) |
| [#5403](https://github.com/CrocSwap/integer-mult-bounds/pull/5403) | Native-map and scalar cancellation | [Source](https://github.com/aliromman/integer-mult-bounds/tree/ff78b33a8c5327199af63325b0d4ee80c65b429d/research/native-gauge-cancellation) | [Success](https://github.com/CrocSwap/integer-mult-bounds/actions/runs/40034681556) |
| [#5411](https://github.com/CrocSwap/integer-mult-bounds/pull/5411) | Same-paid coupled native inverse | [Source](https://github.com/aliromman/integer-mult-bounds/tree/123ffa6aa7c1866ac60c67739841f71d95b3edf7/research/native-coupled-inverse) | [Success](https://github.com/CrocSwap/integer-mult-bounds/actions/runs/40044877623) |
| [#5439](https://github.com/CrocSwap/integer-mult-bounds/pull/5439) | Odd-E8 decoder and separated-lift bound | [Source](https://github.com/aliromman/integer-mult-bounds/tree/e6d50e7ebd0bb1fd4f388930df78df696e56b404/research/odd-e8-direct-sum) | [Success](https://github.com/CrocSwap/integer-mult-bounds/actions/runs/40105392037) |

[RETENTION.json](RETENTION.json) records the exact heads, hosted run IDs,
and the 77 package files. Every file has the same Git blob at its submitted
head and at the canonical main commit above. The main ledger still marks
these packages as partially reviewed. Archive intake and CI success do not
establish scientific acceptance or a complete multiplication theorem.

The separate [review of PR #5341](https://github.com/CrocSwap/integer-mult-bounds/pull/5341#issuecomment-6371583032)
reports an exact reproduction and proof check by James Reed, with Claude
Code assistance. Its scope clarification is now explicit in the PR body:
helper targets stay fixed; only their paid increments are partitioned.
No new validation claim is made by that wording change.

## Research that stopped

[EXPERIMENTS.md](EXPERIMENTS.md) records each tested family and the reason
to stop it. The broad negative screens remain exploratory. A time limit is
not an impossibility proof, and an abstract operator test is not a physical
lower bound. Public findings retain those distinctions.

The last new test is retained in [e-history](e-history/README.md). It uses
the small four-input, four-axis fixture of the original 112-root E adapter.
It does not test the actual 112-cubed or 120-cubed caller. Exact rational
elimination rejects the two missing targets even after all 76 one-axis
queries on the retained values are granted together. A changed spectral
control recovers all four targets. These scripts and source bytes are
preserved to make that finite test inspectable; they do not justify a
separate supplier or kappa claim.

At closeout, the exact history check and the spectral control were replayed
in normal and optimized CPython 3.12.13. All four runs passed. The scientific fields
match the saved receipt; elapsed time is excluded from that comparison.
No large caller, selected cold replay, compiler, outer admission, or formal
kernel was run for this handoff. No independent review of the last test is
claimed. [REPLAY.json](REPLAY.json) records the commands and results.

## Current continuation point

The original producer gap has advanced in related work. The merged
[PR #5464](https://github.com/CrocSwap/integer-mult-bounds/pull/5464), at
`d4b9151a160014370faeb7d080f87a06d88a1acb`, now contains full120 native K
and source-consuming E adapters as well as the retained112 construction.
Its
[current proof](https://github.com/thu-nmrc/integer-mult-bounds/blob/d4b9151a160014370faeb7d080f87a06d88a1acb/research/full120-e8-lowrank-e-native/author/PROOF.md)
and explicit invoice are the source for a new caller task. That work is
credited to THU New Media Research Center and its retained source authors;
it is not a result of this handoff. Its new package was not replayed here.

The negative bank still has the coded endpoint
`Q^-1 E3 E2^-1 E1 D123`. Decoding only the positive bank does not supply
the missing negative Gaussian outputs. The merged
[PR #5508](https://github.com/CrocSwap/integer-mult-bounds/pull/5508), at
`7e6824d560337e2b2e49da8b98727183491f6ecf`, excludes the actual112-root
postdecoder from its stated columnwise ordinary algebra. Paid decoders,
correlated histories, and changed targets are outside that obstruction.
The retained small fixture does not extend it to the full caller.

## Open next questions

1. Construct a paid negative decoder, or change the caller and its target
inventory. Use the current full120 source and retain its exact order of
Q and E. First state the full optimistic bill. Continue only if the bill
can exceed the selected result. Then check every original return,
helper, adapter, idle, COPY/CLEAR, route, precision, and storage cost.
2. If using the old coupled provider, change the source or the caller
premise. The fixed-source guard and history tests below give specific
stop rules. They do not exclude all other protocols.
3. For a proof review of a delivered package, use its exact head and its
recorded open obligations. Do not treat an archival merge as an audit
of the full theorem.

All reservations made by this investigation are released. Original claim
posts were updated as well as their completion replies. Dillon's
[answer to the control-definition request](https://github.com/CrocSwap/integer-mult-bounds/discussions/4930#discussioncomment-19161071)
is recorded as answered. No answer or reserved work is owed to this run.
These are continuation questions, not tasks assigned to another contributor.

## Evidence retention and credit

The complete local record contains the three original work directories,
their checkpoint notes, source inputs, scripts, receipts, failed attempts,
and publication checks. A file inventory and a verified archive preserve
them. A second pass through this run's tool history checked for missed
temporary inputs and claim posts. The local record is broader than this
public handoff. It does not certify every exploratory script or claim that
every search was rerun at closeout.

Credit remains with the source authors named by each linked package and
finding. In particular, retain Cao Yifan's native word; Dillon's provider,
normalizers and clarification; Dugongue's geometry; the compiler and
precision lineage; the THU New Media Research Center E adapters; and the
attributed parity-projector result used in the fixed-pattern finding.
Mark's [separate product-plane reproduction](https://github.com/CrocSwap/integer-mult-bounds/discussions/1587#discussioncomment-19168830)
confirms the two factor tariffs. It also explains why rank alone gives no
tariff floor. Neither factor is claimed optimal.

Prepared by Ali Romman with substantial OpenAI Codex assistance. No
subagents were used in this closeout. Apache-2.0. No other contributor's
PR was edited, merged, or closed by this closeout.
Loading