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
42 changes: 42 additions & 0 deletions research/p104-paid-interior-bases/NOTICE
Original file line number Diff line number Diff line change
@@ -0,0 +1,42 @@
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.
26 changes: 26 additions & 0 deletions research/p104-paid-interior-bases/PRICING-EDITS.json
Original file line number Diff line number Diff line change
@@ -0,0 +1,26 @@
[
[
"Q('.00310676333'), Q('.003106763335')",
"Q('.00310815849'), Q('.003108158495')"
],
[
"PREVIOUS = Q('.00310644469')",
"PREVIOUS = Q('.00310676333')"
],
[
"== 66474",
"== 66402"
],
[
"'kappa_decimal':'.00310676333'",
"'kappa_decimal':'.00310815849'"
],
[
"previous_open_pr5651",
"previous_open_pr5660"
],
[
"Ali Romman PR5644. The four-gate change is by Ali Romman with Codex assistance.",
"Ali Romman PR5644. The four-gate change is by Ali Romman with Codex assistance.\nThe four interior basis changes are by Ali Romman with Codex assistance."
]
]
111 changes: 111 additions & 0 deletions research/p104-paid-interior-bases/PROOF.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,111 @@
# Exact construction and complete price

## 1. One temporary basis

Let `a` and `b` be distinct public registers. Set `R = I + c E_ab` for a
nonzero dyadic scalar `c`. Then `R^-1 = I - c E_ab`.

Choose two internal cuts where both registers can reach one common chart.
Insert `R` at the first cut. Between the cuts, replace each scalar block
`M_j` by `R M_j R^-1`. Insert `R^-1` at the second cut. The complete interval
is unchanged:

```text
R^-1 (R M_k R^-1) ... (R M_1 R^-1) R = M_k ... M_1.
```

A block that does not use either register commutes with `R`. The same is
true for each owned initialization and erasure, since `a` and `b` are public.
The checker requires one certificate for every affected block. It checks
the exact old event list and every rational matrix coefficient. It also
rejects extra certificates. Thus this is an identity for arbitrary inputs,
not a test on selected vectors.

Each scalar block has one physical chart. The compiler moves each operand
to that chart before it is used. At every step, its physical value is the
declared native chart applied to its logical value. A scalar basis change
at a common chart therefore implements the same rational matrix `R`.
Native transport changes the chart and preserves the logical value.
The final basis removal restores the original logical interval.

## 2. Fixed witnesses

The four certificates are applied in the order in WITNESS.json. Block
indices refer to the complete word produced by the preceding step.

| Public pair | Coefficient | Local paid delta |
| --- | ---: | --- |
| 438, 733 | -4 | `-H1-H2+H3` |
| 119, 125 | -1 | `-2H1+H2` |
| 125, 128 | -1 | `-2H1+H2` |
| 128, 139 | -1 | `-2H1+H2` |

There are eight changed scalar blocks. Each certificate gives the full
before and after lists. The new contact graph permits four fewer local
native calls. The transport builder includes both meeting charts and all
remaining scalar contacts. It does not assign free movement to a basis
operation. All public paths remain nested.

The separate endpoint checks cover 743 arbitrary public inputs and nine
owned register slots. There are eleven owned lifetimes and peak occupancy
two. Every owned instruction and transport stays in the same order. All
535 productive helper targets and all public boundary charts stay fixed.

## 3. Native and caller binding

The retained compiler lowers each chart change to a signed Gaussian-dyadic
recipe. A separate interpreter checks every signed Pauli image and the
global phase. It checks 2,918 distinct maps and 208 main endpoint adapters.
The helper normalizers retain the exact productive target profile.

The p16/a7 caller is unchanged. Its physical binding checks all 7,488
contacts and all 53,248 central-square columns. It has 18 full forward
calls, 15 thin query shears and 3,120 paid whole-idle jobs. The 3,328 main
targets and 9,630 helper targets give 12,958 productive fields.

These exact finite identities use the retained factorwise tensor rule.
They do not add a new analytic or all-size admission assumption. The five
existing framework assumptions and helper endpoint contract remain.

## 4. Complete cost

For `H_r(t) = r^t` and `0 < t < 1`, the paid change per local call is

```text
delta P = -7 H1 + 2 H2 + H3,
delta T = 0.
```

Strict concavity gives `1+2^t>3^t` and `2>2^t`. Thus the complete signed
moment increases by `18 * (7 - 2*2^t - 3^t) > 0` throughout that interval.
The new bill has 66,402 paid calls and rank surplus 760. The local paid rank
mass remains 5,986; the helper rank mass remains 4,266. Their difference is
the 1,664 main progress plus 56 owned descent debt.

Two rational interval engines verify the full moment and put its unique
raw crossing in `(0.00310815850980, 0.00310815850981)`. The claim is
`0.00310815849`, and the tested saving is `0.003108158495`.

All scalar changes are priced. The local scalar allowance is 220,290 and
the route allowance is 5,167,000. The retained complete raw ordinary bound
is 184,647,552, below its 1,000,000,000 budget. The power-1024 compiler,
cutoff 17, floor 18, tail parameter 1/5, reserve 10^-12 and decay 39/40 are
unchanged. All 54 strict and four functional outer conditions pass.

## 5. Verification limits

The full source archive is authenticated before any retained module is
loaded. The nested predecessor archive is authenticated again before its
local or native checker is loaded. The new checker verifies the source
word and every intermediate complete word. Its separate dense transpose
engine checks 565,504 entries.

False basis signs, false block coefficients, a missing block certificate,
a missing native transport, a missing owned erasure, an unpaid query and
extra target credit are rejected. The inherited phase, caller and price
controls also run.

The huge uniform word and rounded multiplier are algorithm-defined and
are not materialized. These are author checks, not independent scientific
acceptance or a proof-kernel theorem. The numerical improvement is small.
No large gain or global optimum is established.
81 changes: 81 additions & 0 deletions research/p104-paid-interior-bases/README.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,81 @@
# Paid interior bases for the P104 word

Complete conditional kappa is **0.00310815849**, a gain of **0.00000139516**
(about **0.04491%**) over
[PR #5660](https://github.com/CrocSwap/integer-mult-bounds/pull/5660).
This is a small gain. A large improvement and a global optimum remain open.

Four temporary changes of the public register basis remove paid contacts.
The basis changes occur at internal meeting charts. All transports to those
charts are included. The source and output interfaces stay fixed.
The complete local delta is `-7 H1 + 2 H2 + H3`. Across 18 full calls, this
removes 72 paid requests. The complete bill has **66,402 paid calls**,
**12,958 productive targets**, and rank surplus **760**.

[PROOF.md](PROOF.md) states the identity and the complete cost.
[WITNESS.json](WITNESS.json) contains all eight scalar block certificates.
[RESULT.json](RESULT.json) is the full receipt.
[SOURCES.json](SOURCES.json) pins every consumed source.

## Replay

Python 3.12 and the standard library suffice. From the repository root:

```sh
python3.12 -B scripts/research_package.py --package research/p104-paid-interior-bases
```

The hook runs normal and optimized checks. The expected final marker is:

```text
PASS_P104_PAID_INTERIOR_BASES_KAPPA_00310815849
```

The direct commands also work in a source export with no Git or network:

```sh
python3.12 -I -B research/p104-paid-interior-bases/verify.py
python3.12 -I -O -B research/p104-paid-interior-bases/verify.py
```

The checks authenticate the 23-file PR5660 export and its 36-file predecessor
export before source imports. They check every exact block identity, all
565,504 transpose coefficients, 2,918 signed native maps, the physical caller,
both rational moment engines, all compiler costs, and all 58 outer conditions.
See [VALIDATION.md](VALIDATION.md) for the exact checks that ran.

## Scope and limits

The five inherited framework assumptions and the helper endpoint contract
remain. The huge uniform program and complete rounded multiplier are not
materialized. Author checks are not independent scientific acceptance.
The selected frontier is unchanged.

The source is PR5660 at `3b0b4f37d5005f381ed9cfcccebfbe1ab35c71aa`.
The branch starts from canonical main `df0137efd35e11b4445734bf612bdd4c7ebc330c`,
priority revision 7. The complete source is retained, including all notices.

## Open next questions

1. Can an independent reviewer confirm the block identities and full bill?
2. Can a larger temporary basis or a different scalar schedule give a large
gain after all meeting transports and ordinary work are paid?
3. Can one of the inherited framework assumptions be resolved?

The finite discovery searches do not prove a general obstruction. Further
work must check live Discussions and pin its complete input word.

## Credit

[PR #5660](https://github.com/CrocSwap/integer-mult-bounds/pull/5660) supplies
the complete input and the four-gate contact rewrite. It retains zxv's
[PR #5643](https://github.com/CrocSwap/integer-mult-bounds/pull/5643) A2 change,
the [PR #5644](https://github.com/CrocSwap/integer-mult-bounds/pull/5644) helper
relations, and the [PR #5650](https://github.com/CrocSwap/integer-mult-bounds/pull/5650)
composition. The source notices retain the P104, side-correction, helper,
native, caller, productive-recursion, interval and compiler authors.
The elementary change-of-basis identity is not claimed as a new theorem.

The paid interior construction and this verification package are by
Ali Romman with substantial OpenAI Codex assistance. No subagents.
Apache-2.0. See [NOTICE](NOTICE) for the full retained attribution.
Loading
Loading