Skip to content
Open
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
61 changes: 61 additions & 0 deletions research/p104-receiver-single-send/PROOF.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,61 @@
# Sharp direct repair cost at receiver 127

## Cut and price

Fix the rational word with SHA-256 c447b6d5ba956fbe40834fe3c1bc69af2f225736f1b97ecd84b57bab9dd283cb on 1,289 independent rational coordinates. Its 5,709 paid transfers and 1,289 terminal rules run on 1,103 fixed charts. The old path is chart $278\to288\to127$ through slots 95914 and 88. Its grades are 2 and 5. Remove both transfers. Keep the original initial stocks, all other paid rows, all other routes, all other chart assignments, and the exterior terminal rules. The 901 surviving deposits at receiver 127 may be combined freely. It has no initial coordinates and no outgoing paid transfer. Slot 95914 had only one reader, slot 88. Slot 88 had only two readers, both receiver terminals. Thus removing the path leaves every surviving paid stock value equal to its original value. [Pinned incumbent](https://github.com/CrocSwap/integer-mult-bounds/tree/89a67c404226d51028e135de2b9169e9020d57af/research/p104-iterated-returns/continuation) · [Predecessor stock contract](https://github.com/CrocSwap/integer-mult-bounds/blob/0893abefcc894991b67b2315f7e85e0dc4d58c5e/research/p104-paid88-stock-cones/STOCK.md)

Allow any finite number of new direct transmissions from any original chart to receiver 127. The calculation charges only their paid grades and grants all local combining and receiving rules for free. Even the lower-bound relaxation lets each donor see every surviving arrival at its chart over the whole word. In its deterministic adaptive version the receiver may announce its entire transcript freely and choose the next donor from that transcript. Donors receive no other extra input. The upper bound uses a fixed chronological, rational-linear word with two sends at the original slots. No lower bound here requires speculative information about a later operator or the cost of a receiver addition.

Chart 127 spans the whole nine-dimensional binary frame space. Every original chart of dimension $r$ is contained in that space, so its direct send has grade $9-r$. For any nonnegative nondecreasing discrete concave tariff $p$ on grades $0,\ldots,9$, with $p(0)=0$, the minimum paid replacement cost is

$$
p(8)+p(1).
$$

The minimum number of direct sends is two. The proof separately establishes impossibility for one send and attains the displayed tariff with two.

## Four inputs separate all senders

Give all unlisted coordinates value zero. Evaluate the original local rational rules on the following four inputs. A chart is live if some initial coordinate or surviving deposit in its full-time stock differs from its value on zero input. Each test has zero on all 901 surviving receiver deposits and value one on terminal 155442. Each requires at least one message from a live chart.

| Test | Nonzero public coordinates | Number of live charts | Property used |
|---|---|---:|---|
| A | $x_{674}=1,\ x_{722}=1$ | 3 | Exactly charts 11, 58 and 278 |
| B | $x_{722}=1,\ x_{780}=-1$ | 293 | Chart 11 is silent and each live chart has grade at least 1 |
| C | $x_{674}=1,\ x_{825}=-1$ | 295 | Chart 58 is silent and each live chart has grade at least 1 |
| D | $x_{674}=1,\ x_{722}=1,\ x_{1288}=-1/845660160$ | 26 | Chart 278 is silent and each live chart has grade at least 5 |

Test D comes from Clear1nsight's paid-pair obstruction. Tests A–C and the full-chart stock census supply the simultaneous cover here. The forward verifier authenticates the exact compressed source and uncompressed payload, checks each rule against the sender's chronological stock, and propagates the four inputs in exact rational arithmetic. The earlier and later path slots are excluded from the census of surviving arrivals. The verifier also computes all binary frame ranks independently. [Original input D](https://github.com/CrocSwap/integer-mult-bounds/blob/0893abefcc894991b67b2315f7e85e0dc4d58c5e/research/p104-paid88-stock-cones/verify_stock.py) · [Forward verifier](verify_cut.py)

For any successful fixed strategy the chosen senders must intersect all four live sets. This applies also to the zero-input branch of a deterministic adaptive strategy. If its branch misses a live set, apply the corresponding test. The local input and prior transcript at every queried chart then coincide with their values on zero. The receiver takes the same branch and gets the same answer, although its required terminal changes from zero to one. Arbitrary deterministic encodings and nonlinear receiver postprocessing obey this implication.

Test A forces a donor from $\{11,58,278\}$. If chart 11 is used, its grade is 8 and B requires another sender at grade at least 1. If chart 58 is used, C does the same. Otherwise chart 278 is used at grade 7 and D requires another sender at grade at least 5. No individual chart passes all tests. Concavity and monotonicity yield $p(8)-p(7)\le p(2)-p(1)\le p(5)-p(1)$. Every allowed direct strategy therefore costs at least $p(8)+p(1)$. This bound includes arbitrary extra grade-zero messages.

## Chronological construction and exact operator

At slot 95914 send the initial $x_{674}$ from chart 11 directly to chart 127. At slot 88 send stock atom 101539 from chart 138 directly to chart 127. Its old deposit occurs at time 3396, strictly before the replacement send at time 4067, and its full public-coordinate expansion is $x_{780}$. In the original word it is the half difference of 98143 and 100323. Their expansions are $x_{825}+x_{780}$ and $x_{825}-x_{780}$. The new pair has grades 8 and 1. On predecessor witness D the first donor sees a nonzero value. The two senders together recover the difference required on all public inputs.

Let $y_1$ and $y_2$ be the original terminals 155442 and 155490, and let $t$ be the old outlet 88. Both rows have coefficient one on $t$. Write $r_i=y_i-t$ for the original rows with that local term removed. Reverse substitution gives $y_1=x_{674}-x_{780}$. The receiver returns
$$
y'_1=x_{674}-x_{780},\qquad
y'_2=r_2-r_1+x_{674}-x_{780}.
$$
Every term in this expression is available under the literal schedule. In particular, terminal 155490 is actually installed *before* 155442, so its evaluator uses the two arriving atoms and the old deposit rows directly. The implementation checks availability of every term at both installations. It changes exactly two transfers and two terminal forms. The other 1,287 terminals have no dependency path through a changed node. Independent backward substitution verifies both changed outputs as linear forms in all 1,289 coordinates. Omitting the second receiver recoding or reversing the two source signs is detected. [Constructive verifier](verify_operator.py)

## Price at the signed target

The incumbent tariff at $\kappa=0.00387981257853$ is $p(d)=d^{1-\kappa}$. Keeping the productive fields and other paid transfers fixed, the optimal direct replacement increases the 48-call paid cost by
$$
48\big(8^{1-\kappa}+1-2^{1-\kappa}-5^{1-\kappa}\big)
\ \in\ (94.666202960674077138503464,\ 94.666202960674077138503465).
$$
The enrolled 48-call invoice already needs more than 0.379442 signed units of saving at this target. Even granting all local recoding for free, the best direct replacement increases the paid bill. This rules out reaching that target by this family of changes alone. Additional modified paid rules, cooperative intermediate exchanges or altered productive credits change the optimization problem. [Frozen incumbent tariff and target shortfall](https://github.com/CrocSwap/integer-mult-bounds/blob/0893abefcc894991b67b2315f7e85e0dc4d58c5e/research/p104-paid88-stock-cones/cones/SOURCE.json)

For power tariffs $p(d)=d^s$, $0<s\le1$, this exact cost has a unique crossover relative to the old path. The ratio
$$
\frac{8^s-5^s}{2^s-1} =
\frac{\int_5^8 x^{s-1}\,dx}{\int_1^2 x^{s-1}\,dx}
$$
strictly increases with $s$. Indeed, raising $s$ by $h>0$ multiplies the numerator by at least $5^h$ and the denominator by at most $2^h$. At $s=0$ the continuous limiting ratio is $\log(8/5)/\log 2<1$, and at $s=1$ it is 3. The unique positive equality lies between 0.25985761 and 0.25985762. The direct optimum is dearer above that crossover. The price verifier bounds the logarithm by a rational atanh series and its geometric tail, then encloses the exponential between alternating Taylor sums. All published price endpoints are obtained without floating-point arithmetic.

Prepared by apoursh with substantial OpenAI Codex assistance. Apache-2.0.
32 changes: 32 additions & 0 deletions research/p104-receiver-single-send/README.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,32 @@
# Exact price of replacing P104 paid slot 88 with direct sends

On the submitted rational P104 word from #7786, remove the route through chart 288 at paid slots 95914 and 88. Allow any of the 1,103 original charts to send directly to receiver 127 while preserving all other transfers and permitting arbitrary receiver recoding. The least number of direct sends is **two**. For every nonnegative, nondecreasing concave grade tariff with zero cost at grade zero, the exact optimal paid cost is **one grade-eight send and one grade-one send**. [Proof and model](PROOF.md)

At the enrolled target κ = 0.00387981257853, the incumbent route pays grades two and five. Across the 48 rational calls, the best direct solution **increases** the paid bill by 94.666202960674077138503464–94.666202960674077138503465 signed units. The incumbent already needs 0.379442 additional units of saving. This is a sharp obstruction for direct replacement of a fixed cut. A physical implementation would also pay for local restocking and terminal arithmetic.

| What was checked | Result |
|---|---|
| Four exact sparse inputs on the complete word | All 901 surviving receiver deposits vanish. The required output is one. |
| Every donor with all-time stock | The first test is visible only at charts 11, 58 and 278. Each requires a different chart on another test. |
| Constructive replacement at the original times | Chart 11 sends public coordinate 674. Chart 138 sends an already available exact copy of coordinate 780. |
| Source and replacement operators | Two rewritten terminal forms agree on all 1,289 independent coordinates. The other 1,287 outputs have no changed dependency. |

Clear1nsight's #7795 asked for an additional live observation on a three-coordinate witness after six proposed paid-pair recodings failed. Here chart 11 supplies one, and combining it with chart 138's live atom actually reconstructs the required rows. The four-witness argument prices the optimum across every possible set of fixed direct donors. Changes to other paid rows and inter-donor routing remain open. Clear1nsight's concurrent #7804 tests 126 two-ingress choices at adjoint chart 3 and a changed exterior. Its source and receiver boundary differ from the direct primal chart-127 cut studied here. [Predecessor handoff](https://github.com/CrocSwap/integer-mult-bounds/pull/7795) · [Parallel result #7804](https://github.com/CrocSwap/integer-mult-bounds/pull/7804) · [Coordination claim](https://github.com/CrocSwap/integer-mult-bounds/discussions/7111#discussioncomment-19754778)

## Reproduce

From the repository root, run the registered package with CPython 3.12.

~~~sh
python3.12 -B scripts/research_package.py --package research/p104-receiver-single-send
~~~

The standalone checks are standard-library Python. The forward check propagates the four witnesses, verifies all chart ranks, and independently encloses the signed tariff. The second check constructs the legal two-send operator, validates all chronological sources, and verifies both changed outputs by reverse substitution. The immutable compressed word and its uncompressed payload are authenticated on each run. [Forward check](verify_cut.py) · [Adjoint check](verify_operator.py) · [Source pins](SOURCE-PINS.json)

## Open next questions

Can a change to an existing ingress at chart 278 combine the useful observations before the final direct leg at sufficiently low cost? Start by finding an ingress operator that carries the weight of coordinate 674 on test A and separates test D while preserving its original downstream readers. Price the altered ingress, the final send, and every required restoration against the incumbent 48-call invoice. Existing exclusions for six named pairs and the present fixed-stock optimum constrain different classes of changes.

## Credit

Clear1nsight supplied the #7786 rational word, the paid-slot-88 handoff and test D in #7795, building on Ali Romman's #7776 signed framework, the #7774 and #7781 lineages, and the contributors recorded in the retained notice. The other three sparse inputs, all-chart proof, exact attaining recoding and price comparison were prepared by apoursh with substantial OpenAI Codex assistance. New work is Apache-2.0. The project's selected conditional multiplication bound remains unchanged. [Incumbent source and contributor history](https://github.com/CrocSwap/integer-mult-bounds/tree/89a67c404226d51028e135de2b9169e9020d57af/research/p104-iterated-returns/continuation) · [Retained source notice](retained/PR7795-NOTICE) · [Original word notice](retained/PR7786-NOTICE) · [Source license](retained/LICENSE)
49 changes: 49 additions & 0 deletions research/p104-receiver-single-send/SOURCE-PINS.json
Original file line number Diff line number Diff line change
@@ -0,0 +1,49 @@
{
"artifacts": [
{
"local": "retained/WORD.json.gz",
"source": 7786,
"head": "89a67c404226d51028e135de2b9169e9020d57af",
"original_path": "research/p104-iterated-returns/continuation/WORD.json.gz",
"bytes": 614180,
"git_blob": "337a9d4f8b36f3bd4acc5651810e4e846e287435",
"sha256": "fde0f7eaa3749d0934a9fc0d6320074f021cd722ed6753f3a59a62fb5b221949",
"uncompressed_sha256": "c447b6d5ba956fbe40834fe3c1bc69af2f225736f1b97ecd84b57bab9dd283cb"
},
{
"local": "retained/PR7795-NOTICE",
"source": 7795,
"head": "0893abefcc894991b67b2315f7e85e0dc4d58c5e",
"original_path": "research/p104-paid88-stock-cones/NOTICE",
"bytes": 1432,
"git_blob": "6cbfdc843fd7cd3227c7bd7925c891499d93098d",
"sha256": "8df09db925c14a4ec17e578455dd68b72714e15990cee40936212ce7408132cc"
},
{
"local": "retained/PR7786-NOTICE",
"source": 7786,
"head": "89a67c404226d51028e135de2b9169e9020d57af",
"original_path": "research/p104-iterated-returns/NOTICE",
"git_blob": "3b083891bf59f449b49df024ab7df3feaf6054b8",
"sha256": "4f7d16a4dcaae02de7e7583f7fbb14d5edd19dc4c10b4f47571c145204505844",
"bytes": 1048
},
{
"local": "retained/LICENSE",
"source": 7786,
"head": "89a67c404226d51028e135de2b9169e9020d57af",
"original_path": "research/p104-iterated-returns/LICENSE",
"git_blob": "261eeb9e9f8b2b4b0d119366dda99c6fd7d35c64",
"sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4",
"bytes": 11357
}
],
"tariff": {
"pr": 7795,
"head": "0893abefcc894991b67b2315f7e85e0dc4d58c5e",
"path": "research/p104-paid88-stock-cones/cones/SOURCE.json",
"git_blob": "c0da8ca07dc256f47a980e5ad1a969f7616ab08a",
"sha256": "f5087c4d8aa394049eb0ab08adace0d295009306900f29e9db1a07706a10fba2",
"use": "Tariff and incumbent price context. The replay computes its own rational intervals."
}
}
41 changes: 41 additions & 0 deletions research/p104-receiver-single-send/SOURCES.json
Original file line number Diff line number Diff line change
@@ -0,0 +1,41 @@
{
"schema_version": 1,
"base_commit": "e54523517d0f03e58352e647c403a5af3e525a57",
"files": [
{
"path": "certificates/selected-result.json",
"usage": "provenance",
"git_blob": "75d77b328ea02ded535e7f23aee5b1e55d0bb1a8",
"sha256": "154c3d30455d778fd3343b6b4851f0fa4665f0f7625b0adb3675e2bafbbf6bf1"
},
{
"path": "docs/research/agent-task-catalog.json",
"usage": "provenance",
"git_blob": "3404fbdd5e7529a0674eef26114b5071e4586cec",
"sha256": "ead6cdc9d0af6d58e8a49744d8ec8c7c15e13d2ba93c276c434bcf9b51b93a81"
}
],
"cited_research": [
{
"pr": 7786,
"head": "89a67c404226d51028e135de2b9169e9020d57af",
"author": "Clear1nsight",
"title": "Conditional κ ≥ 0.00387944011068 from four additional rational returns",
"url": "https://github.com/CrocSwap/integer-mult-bounds/pull/7786"
},
{
"pr": 7795,
"head": "0893abefcc894991b67b2315f7e85e0dc4d58c5e",
"author": "Clear1nsight",
"title": "Bound six paid-slot repairs and 73 linked chart cones on #7786",
"url": "https://github.com/CrocSwap/integer-mult-bounds/pull/7795"
},
{
"pr": 7804,
"head": "51e63c4ee7b3bdb05601b453cc1d6838e43bd001",
"author": "Clear1nsight",
"title": "Exclude 126 fixed-port two-ingress repairs at P104 adjoint chart 3",
"url": "https://github.com/CrocSwap/integer-mult-bounds/pull/7804"
}
]
}
Loading
Loading