Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
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
63 changes: 63 additions & 0 deletions research/p104-second-low-rank-block/NOTICE
Original file line number Diff line number Diff line change
@@ -0,0 +1,63 @@
Second rank-four contact block on the complete PR5756 word.
Ali Romman with substantial OpenAI Codex assistance. Apache-2.0.
No subagents. These are author-side checks.

Joel Pulikkan (GamingPuzzled), with Claude Opus 5.5 assistance, supplied the
R22 reframe mechanism and the different block in PR5739/5756.
This package adds a different 102-operation block; it does not claim that mechanism.

The source PR5756 notice follows unchanged.

Paid intermediate meetings and collective helper splits.
Ali Romman, with substantial OpenAI Codex assistance. Apache-2.0.
No subagents. All checks are author-side calculations.

The complete PR5718 notice follows unchanged.

Four paid interior public-basis changes.
Ali Romman, with substantial OpenAI Codex assistance. Apache-2.0.
No subagents.

The following exact notice is retained from PR5660.

New helper-subspace construction, proof and replay:
Ali Romman with substantial OpenAI Codex assistance. Apache-2.0.
The same assistant performed the construction and author-side checks.

Retained contributions:
flyingviper: PR5605 shared relay, PR5612 delayed relay, and the PR5608/PR5613
relay/split compositions.
Scratch-net: PR5606 prepared-helper identities and target debit.
Pierre-Loic Doulcet (hexapode): PR5565 commuting-contact construction, used as the exact source.
Ali Romman: PR5602 fixed boundary relations, PR5601 three-contact frame
change, PR5580 contact step,
and PR5544/5545 complete P104 caller.
Joel Pulikkan (GamingPuzzled): PR5569 priced helper handoff.
DreamingOfClouds: PR5561 free helper endpoints and PR5568 search tools.
thu-nmrc: PR5551 contact improvements.
Gareth Davidson (bitplane): PR2305 source composition.
David Arias: PR2152 coalesced P104 word.
Cinna Davis: PR2219 private reuse and the controller lineage.
Jacob Alan Sussman: productive mixed-target recursion.
zxv and Cao Yifan: exact intervals, compiler and native/pricing lineage.
GamingPuzzled: PR5472 successor handoff, a predecessor to the split concept.
The retained source notices preserve earlier geometry, BIT and analytic credit.

The elementary rank-one basis identity is not claimed as a new general
linear-algebra theorem. The contribution is its explicit independent-target
use on this P104 word and the complete conditional kappa improvement.

New contribution: Ali Romman, with substantial OpenAI Codex assistance.
The four-gate source-version rewrite composes zxv PR5643 and Ali Romman PR5644.
The A2 method retains Chris R/Clear1nsight PR3998 and ToikY PR4019 credit.
The compensated triangle method retains Cinna Davis PR4692 credit.
No subagents were used for this contribution. Apache-2.0.

Concurrent PR5650 by zxv publishes the PR5643/PR5644 composition. PR5651 by
Ali Romman corroborates it. This package credits that released composition
and contributes the further four-gate placement.

New work: distributed scalar updates on an unobserved public field interval, the exact witness and complete conditional price. Prepared for Ali Romman with substantial OpenAI Codex assistance. No subagents. Apache-2.0. The source archive retains all predecessor notices.

Two-boundary main-assisted helper split and early cleanup: flyingviper with substantial
OpenAI Codex assistance, including independent AI review agents. Apache-2.0.
34 changes: 34 additions & 0 deletions research/p104-second-low-rank-block/PLAN.json
Original file line number Diff line number Diff line change
@@ -0,0 +1,34 @@
{
"basis": [
11,
21,
38,
448
],
"old_frame": 267,
"roles": [
57,
81,
92,
224,
225,
227,
229,
234,
742
],
"initial_helper": 742,
"final_helper": 225,
"runs": [
{
"first_scalar": 9858,
"length": 90,
"old_sha256": "2c4d8332eed5d21830748b694d88a830f98f50773e7faddfb65b654800d0e296"
},
{
"first_scalar": 11957,
"length": 12,
"old_sha256": "bcb7d071dbe4b830053892decd7c8b6aef5a26b06a06e010148882871cf07ee2"
}
]
}
26 changes: 26 additions & 0 deletions research/p104-second-low-rank-block/PRICING-EDITS.json
Original file line number Diff line number Diff line change
@@ -0,0 +1,26 @@
[
[
"KAPPA, TESTED = Q('.00311301600'), Q('.003113016005')",
"KAPPA, TESTED = Q('0.00311325989'), Q('0.003113259895')"
],
[
"PREVIOUS = Q('.00311024570')",
"PREVIOUS = Q('.00311313795')"
],
[
"sum(paid.values()) == 66834",
"sum(paid.values()) == 66870"
],
[
"'kappa_decimal':'0.00311301600'",
"'kappa_decimal':'0.00311325989'"
],
[
"'previous_pr5718':PREVIOUS",
"'previous_pr5756':PREVIOUS"
],
[
"Paid intermediate meetings: Ali Romman with substantial OpenAI Codex assistance.",
"Paid intermediate meetings: Ali Romman with substantial OpenAI Codex assistance.\nThe source R22 block is by Joel Pulikkan (GamingPuzzled).\nThe second rank-four block is by Ali Romman with OpenAI Codex assistance."
]
]
91 changes: 91 additions & 0 deletions research/p104-second-low-rank-block/PROOF.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,91 @@
# Proof and full cost

## Exact local change

Let C(F) be the retained native chart map for binary subspace F. A MOVE from F
to G implements `C(G) C(F)^-1`. The source checker factors each map into its
recursive calls and ordinary operations. It checks the signed Pauli action
and the exact global phase. The frame induction then applies at every positive
tensor width. A scalar instruction acts on registers in one common chart.
Changing that chart preserves its abstract scalar action.

The constructor uses the two exact runs in PLAN.json. It keeps their scalar
instructions unchanged and assigns the rank-four chart. It rebuilds all MOVE
instructions from the complete scalar sequence. The only changed public
boundary charts belong to helpers 742 and 225. The source contract permits
these free helper endpoints and charges their resulting target transforms.
No main endpoint changes. No helper input is assumed to be zero.

`verify.py` checks every scalar coefficient on all 767 public inputs and nine
owned coordinates. It separately checks every transpose column in reverse
chronological order. The native replay checks all maps and phases, every main
adapter and every helper normalizer. Thus this check includes the boundary
cost, not just the interior rank-four operations.

The local paid profile is:

```text
rank: 1 2 3 4 5 6 7
count: 2221 454 492 94 173 6 15
```

The local productive helper profile is:

```text
rank: 1 2 4 5 7 8 9
count: 23 5 93 8 6 36 388
```

Compared with the frozen source, the paid delta is `H1+H4-H5` and the helper
delta is `-H1+H4+H5-H8`. Paid rank mass stays 5,987. The scalar allowance stays
224,138. There are eleven owned lifetimes and at most two live owned fields.

## Physical caller and complete price

The unchanged productive controller makes eighteen full local calls. It also
pays fifteen thin queries, each of cost `104 H1`, and fifteen whole idles.
The idle ranks are `8L`: two instances of each L from two through eight and
one instance of L=9. Each instance costs 208 such jobs.

Its main productive targets are `208 H145` and the fifteen idles with rank
increased by one. The full target profile is this main profile plus eighteen
copies of the local helper profile. The full paid profile is eighteen copies
of the local paid profile plus all thin queries and all whole idles.
The physical replay checks each contact and each central-square column.
The full profiles appear literally in RESULT.json. They contain 13,390 targets
and 66,870 paid calls, with rank-mass surplus 760.

For a saving k, set `t=1-k`. The raw condition compares the complete moments
`sum(T[r] r^t)` and `sum(P[r] r^t)`. Both exact rational interval engines give
a positive margin at the tested saving `0.003113259895`. They put the raw zero
in `(0.00311325991155, 0.00311325991156)`. The retained derivative bound is
negative throughout `[0,0.0034]`, which gives the fixed-profile ceiling.
The old complete source profile has a strictly negative upper margin at the
new claimed kappa. This proves a new profile gain, not extra decimal places
from the old profile.

The completion proof uses cutoff 16, floor 17, tail offset 1/7, decay 125/128,
power 1024 and reserve 1e-12. It bounds the whole tail, productive stock and
normalized paid cost. The compiler checks all 54 strict inequalities and
four functional conditions. It includes ordinary scalar operations,
COPY/CLEAR, restoration, native adapters, thin queries, idles and routes.
The inherited BIT saving is 0.0034. The complex cost binds the resulting
conditional multiplication saving `0.00311325989`.

`PRICING-EDITS.json` declares all edits to the archived PR #5751 pricing code:
the claim and tested value, previous claim, full paid-call count, result label
and attribution. The verifier requires those exact edits. It does not weaken
any mathematical price predicate.

## Failure controls and limits

The replay rejects an old helper entry, an old block chart, a missing native
MOVE, a wrong scalar coefficient, an unpaid query, an extra target credit,
a damaged source byte and a missing central translation. The native checker
also rejects the wrong global phase. Explicit exceptions preserve these checks
under Python `-O`.

The finite receipts and inherited frame induction remain conditional on the
five framework assumptions listed in README.md. They do not materialize a
uniform word or rounded multiplier. These are author checks. Independent
review and any selected-frontier admission are separate obligations.
103 changes: 103 additions & 0 deletions research/p104-second-low-rank-block/README.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,103 @@
# Second rank-four block: conditional kappa 0.00311325989

This package gives a complete conditional kappa of **0.00311325989**. It improves
[PR #5756](https://github.com/CrocSwap/integer-mult-bounds/pull/5756)'s claim by
**0.00000012194**. The exact raw crossing lies in
**(0.00311325991155, 0.00311325991156)**. This is a small gain. It does not meet
the donor's request for a much larger improvement.

[PROOF.md](PROOF.md) gives the argument. [RESULT.json](RESULT.json) records the
complete bill and checks. [SOURCES.json](SOURCES.json) pins all retained bytes.
The base is canonical main `de5ec2e7a6074068452f45b1a191b5c72832df4a`, priority
revision 8. The selected frontier is unchanged.

Later [PR #5786](https://github.com/CrocSwap/integer-mult-bounds/pull/5786)
claims the larger bound 0.00312265044 through retained copies and asynchronous
hubs. This package does not supersede that result. Its new block is released
for the separate [source-composition investigation](https://github.com/CrocSwap/integer-mult-bounds/discussions/1166#discussioncomment-19202093); no gain is obtained by adding the published decimal values.

## Change and scope

The exact source is PR #5756 at
`a1b338e5d9271b541f2a198f4fc19cefc565feed`. It contains the paid-meeting changes
from [PR #5751](https://github.com/CrocSwap/integer-mult-bounds/pull/5751) and
Joel Pulikkan's R22 block from
[PR #5739](https://github.com/CrocSwap/integer-mult-bounds/pull/5739).
This package changes a different block on roles
57, 81, 92, 224, 225, 227, 229, 234 and 742.

[PLAN.json](PLAN.json) pins two scalar runs by index, length and SHA256. All 102
operations move from chart 267 of rank eight to `span[11,21,38,448]` of rank
four. Helper 742 starts there; helper 225 ends there. Every scalar coefficient,
scalar order, owned event, main endpoint and port is unchanged. The constructor
rebuilds every native MOVE, including entry and final routes.

Per local call, the paid change is `H1 + H4 - H5`. The helper target change is
`-H1 + H4 + H5 - H8`. Here `Hr` denotes a recursive job of rank r, not an
array-volume exponent. The complete caller has 13,390 productive fields,
66,870 paid calls and rank-mass surplus 760. A count reduction alone is not the
source of the gain: the complete signed moment improves.

The exact raw ceiling applies only to this fixed full target/paid profile.
It leaves between `2.155e-11` and `2.156e-11` above the claim. Further useful
progress must change that profile. No optimum over other words is claimed.

## Replay

Python 3.12 and the standard library are sufficient. From a checkout:

```sh
python3.12 -B scripts/research_package.py --package research/p104-second-low-rank-block
```

The registered commands use ordinary Python and `-O`, in sequence. Each ends:

```text
PASS_P104_SECOND_RANK4_BLOCK_KAPPA_00311325989
```

A source export needs no Git, network or solver. In its package directory:

```sh
python3.12 -I -B verify.py
python3.12 -I -O -B verify.py
```

The source word is frozen, not regenerated from the PR #5756 search. The new
word is regenerated from that frozen word. All 602,176 scalar coefficients and
a separate dense transpose replay are checked. The replay also checks 2,934
native maps with their absolute phases, 208 main adapters, 1,118 helper
normalizers, 7,488 physical contacts and 53,248 central-square columns. Two
rational moment engines check completion and the full bill. All ordinary
COPY/CLEAR, scalar, restoration and route charges remain in the inherited
compiler proof. See [VALIDATION.md](VALIDATION.md) for the run record.

## Conditions and limits

The five inherited framework assumptions remain: `native_all_columns`,
`owned_erasure`, `compiler`, `precision_tape` and `analytic_recovery`.
This is a complete conditional price in that framework. The uniform word and
rounded integer multiplier are not materialized. No independent human review,
proof-kernel acceptance or selected-frontier change is claimed.

## Open next questions

[TASK.json](TASK.json) gives the next questions and their first checks. A larger
gain needs a new full profile. Independent review must still assess this exact
source closure, word and price. Check current Discussions before taking a new
scalar, helper or caller scope.

## Credit

Ali Romman with substantial OpenAI Codex assistance supplied this second block,
its constructor and this replay package. No subagents were used. Apache-2.0.
Joel Pulikkan (GamingPuzzled), with Claude Opus 5.5 assistance, supplied the R22
reframe mechanism and the different block retained from PR #5756. We do not
claim that mechanism as new.

The retained PR #5751 source supplies the exact local, native, caller and price
checks. Its source chain includes flyingviper's distinct-boundary work in
[PR #5718](https://github.com/CrocSwap/integer-mult-bounds/pull/5718), the earlier
P104 helper work, and the controller, compiler and analytic work.
[NOTICE](NOTICE) preserves their notices. Exact source commits, blob IDs and
SHA256 hashes are in SOURCES.json. The nested archives are unchanged.
Loading
Loading