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
172 changes: 172 additions & 0 deletions research/selected-control-reservations/PROOF.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,172 @@
# Selected raw145 routes inside complete compact reservations

This is a source-bound refinement of an existing layout argument. It is not a
new compact-control primitive, a physical execution, or a final-template proof.
The source plan has SHA256
`a02e79fb4615f4f987e3b3617c2b42eea2c1d7415cc66fa702a8b0cd28e6387e`.
The current selected value remains conditional κ=0.003027288235.

## Proposition and quantified boundary

Fix the authenticated `WHOLECALLER-PLAN-v1` with ambient145. Let integers
`d>=D>0`, `K,G,f>=1`, `q0>=1` and `lead>=0` be given. Put

```
H=dG, qF=ceil(2H/K), qB=ceil(H/K),
o=(q0+qF+lead)K, z=(D-qB)K.
```

Assume `(f-1)G<=H` and `o+145fK<=z`. Assume the current stream is a complete
Cartesian product of its row index, the front/back fields, all active and
spectator coordinates, and its record suffix. The source compact-control
primitive, including repair and restoration, is an inherited hypothesis.

Then every literal main/helper/IDLE coordinate route in this plan has the
following placement: source axis `a`, column `j`, selected position `rho`
is the address bit

```
phi(a,j,rho)=o+(af+j)K+rho,
0<=a<145, 0<=j<f, 0<=rho<K.
```

The 119 routes, the 18 embedded local9-planes, and the 16 central translations
use no row or reserved-control bit. For each fixed rho they induce exactly
the plan's columnwise coordinate operation, preserving the other bits of each
K-bit chunk, and restore each entire borrowed compact field. Their statements
hold for every f in the domain, not just the finite f controls. No assertion
about the Gaussian/scalar operator endpoint is newly proved here.

For an IDLE of rank r, its literal `pre_inverseJ` puts exactly its recorded
`active_axes` into the canonical `rf`-chunk child prefix. The other
`(145-r)f` chunks remain complete spectators, followed by any outer active
remainder and prior right spectators. Front/back controls stay outside this
reassociation. Reassembly followed by the literal post-J restores the parent's
coordinate order. This is the address ABI for an assumed complete child
endpoint; it does not execute or price the child.

## Source directions, not an interchangeable J convention

Represent a coordinate word by its pullback list: after swaps, output slot a
contains the input slot recorded at list entry a. Direct composition of every
literal word verifies:

| Family | Number | Pre pullback | Post pullback | Literal swaps |
|---|---:|---|---|---:|
| main targets |32|inverse_permutation|permutation|3,776|
| helper targets |72|inverse_permutation|permutation|10,116|
| IDLE jobs |15|permutation|inverse_permutation|2,856|

In all cases `permutation[:rank]=active_axes`. Thus IDLE's prefix after its pre
word contains its active axes; treating IDLE like a main target reverses the
required direction. Main/helper pre/post words surround different source
boundaries; we do not reinterpret them as IDLE child brackets. Each permutation
is a bijection on all145 axes, and its post word is the inverse of its pre word.
All119 routes have empty affine translation and zero absolute phase. The16
negative-Q main banks separately require the plan's post-LEFT axis0 translation.

The18 local embeddings share axis0; their eight private axes partition1..144.
Every LOCAL points to its stage's exact9-axis list. The source contract declares
that same list on all649 local fields. Our check binds the list and declaration;
it does not rerun the649-column provider algebra or the independent global
integer-ID alpha-renaming join.

## Interval proof and complete dirty fields

The row bits occupy `[0,q0K)`. The two complete front fields occupy
`[q0K,q0K+H)` and `[q0K+H,q0K+2H)`. Their right endpoint is at most
`(q0+qF)K<=o`. The complete back field is `[z,z+H)`, contained in the last
qB chunks. The raw block is `[o,o+145fK)`, contained in `[o,z)`.
These half-open intervals are pairwise disjoint; zero slack is permitted.

For each source axis a, all its selected bits and their unselected guard bits
lie in `[o+afK,o+(a+1)fK)`. Distinct source axes have disjoint slots. Every
source SWAP_BITS is three ordered XOR row additions on two such complete slots.
For each addition the front fields precede both slots and the back field
follows both, in either source order. Carving `(f-1)G` bits from each H-bit
field is legal; all unused bits are spectators. The credited compact proposition
therefore supplies the exact selected-bit addition and full dirty restoration.
At f=1 its elementary two-bit branch is used. Three additions give the swap.
Composition gives each literal route; fixed-rho bit positions are preserved.
An axis0 translation also lies in the same active interval. Local9-plane
restrictions select nine disjoint full slots and leave every other slot present.

The executable interval checks use endpoints to verify whole ranges, including
one f=10^30 case without allocating an array of that size. Their all-f force is
the preceding argument. The separately implemented lifted-word controls expand
all coordinates only for f=1,2,3.

## Whole-row split and actual IDLE child reassociation

Let a current row range have size R, with `R=L*U`. Splitting an entire row by
`r=L*g+w` gives a bijection

```
{0,...,R-1} × Omega <-> {0,...,L-1} × {0,...,U-1} × Omega,
```

where Omega is the entire within-row rectangle, including both front fields,
active slots, back field, spectators and record suffix. No compact or active
bit participates in the row quotient. Consequently each role receives all of
Omega; address completeness is independent of the payload's dirty values.
Repeated legal divisions use mixed-radix row digits, and the inverse formula
restores every original row. A parent/child packet may be correlated with other
owners; this does not change its complete address set.

For each actual IDLE permutation p, the child prefix labels are
`(p[a],j)` for `a<r`, and the complement labels are `(p[a],j)` for `r<=a<145`.
The two lists partition the complete145f labels. Interpreting the second list
as right spectators is a Cartesian reassociation, not a projection or a
reallocation of independent coordinates. The parent controls remain complete
because the primitive restored them before child entry. An assumed child that
preserves spectators and returns its full packet therefore permits the inverse
route on return. The checker verifies all15 actual lists at f=1,2,7, including
complete complementary labels; a dropped spectator is rejected.

If one pads a row count R once to a multiple of a stock product P with R>=P,
then `R'=ceil(R/P)P<2R`. Added rows contain the whole Omega and zero payload;
no partial front/back field is padded. They can be removed only after the
assumed complete linear invocation returns each zero row to zero. Our checker
tests mixed-radix boundary values, including huge integers, but neither assigns
final P nor proves the mathematical endpoint that makes padded rows deletable.
The generic bijection is predecessor work; the contribution is its explicit
join to these source routes and their actual child complements.

## Exact remaining final-compiler obligation

This result concerns the raw145-axis plan, not the exponentially larger
compiled program. `V16-APPLICATION-PROOF-v3` §§5–6 constructs a finite template
with enlarged N,W,P,O and source-style stock products. Its §4 raw145 route bill
cannot supply the final template's actual coordinate injection or row divisor.
In particular10928 is the raw productive count, not a certified final W;
418 is raw private inventory, not a certified final O. The numerical row controls
in this package using10928 and419 are arbitrary-divisor controls only.

To promote this check to that program, supply a source-derived occurrence map
for every chronological product/completion/split/compiler wrapper. It must bind:

1. its actual final-template active slots and rank-child complement;
2. its complete q0/qF/qB ranges, using final W/O/N and the full simultaneous
stock product, with each whole-row divisor legal;
3. the current parent/child descriptors and inverse return placement;
4. the restored-control and spectator premises at every call boundary.

A lazy source-derived generator plus an induction/certificate may suffice;
materializing the astronomical template is not demanded. This package supplies
neither that generator nor its proof. It does not refute the selected written
specialization, whose retained assumptions remain unchanged. It supplies a
reusable raw occurrence lemma and identifies exactly where source binding stops.

## Provenance and coverage

Douglas Colkitt supplies compact reservations, dirty repair and row splitting.
John-Michael Smith, Hanson Sun and Aaron Wang supply retained guard/repair audits.
Jose Manuel supplies PR3055's encoded Compact and the older73-axis physical
controller/row-return implementation. Chris R / Clear1nsight and the selected
PR5413 contributor chain supply the actual145-axis plan and precision/compiler
specialization; all original plan/native/provider credits remain theirs.
PR5445 motivates the actual reservation binding; its first proposed raw-rotation
follow-up is already covered by the retained packages and is not repeated.

New source binding and this scoped proof: JoyboyBrian, with substantial OpenAI
Codex assistance; Apache-2.0. No independent human or proof-kernel acceptance.
74 changes: 74 additions & 0 deletions research/selected-control-reservations/README.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,74 @@
# Selected5413 control-reservation binding

**Scoped positive raw-caller binding; final compiled-template placement remains
open. No new κ.** The selected145-axis plan's119 main/helper/IDLE routes,
16,748 literal swaps,18 local planes and16 central translations fit complete
front/back compact reservations under an explicit parametric address map.

[PROOF.md](PROOF.md) proves the source-bound implication for every positive
column width in the stated complete-field domain. Main/helper pre words and
IDLE pre words use opposite coordinate directions; the verifier retains this
difference. It also binds the15 actual IDLE child prefixes and their full
spectator complements to whole-row division/reassembly.

This does not execute the compact primitive, re-prove its guard, or rerun the
selected Gaussian identity. Those already have substantial retained coverage.
It does not infer final compiler row stock from raw width10928/private count418.
The [precise final-compiler obligation](PROOF.md#exact-remaining-final-compiler-obligation)
requires the expanded occurrence injections and actual W/O/N/stock product.

## Reproduce

Python3.12+, standard library only:

```sh
python3 -I -B research/selected-control-reservations/verify.py
python3 -I -O -B research/selected-control-reservations/verify.py
python3.12 -B scripts/research_package.py --package research/selected-control-reservations
```

Direct replay works without Git or network and authenticates every declared
source input. [RESULT.json](RESULT.json) records the exact census, six interval
families (including f=10^30 without materialization), literal lifted-coordinate
checks,45 IDLE child cases, whole-row boundary controls and eight rejecting
semantic mutations. Source-byte and portable replay controls are recorded separately in
[VALIDATION.json](VALIDATION.json). The fixed q0=11, G=3, K=64 interval fixtures are synthetic address controls,
not production row-stock or physical-tariff instances. Finite executions corroborate the written
quantified proposition; they are not proof-assistant verification.

## Scope and prior coverage

Base32890bc5de034ab8fc93ac88ead7279a6a416d04, priority7; selected conditional
κ=.003027288235. Source pins are in [SOURCES.json](SOURCES.json). Canonical
investigation: [Discussion5477](https://github.com/CrocSwap/integer-mult-bounds/discussions/5477).
The complete fresh overlap snapshot contained342 open PRs and4736 Discussion
records; root coordination owns publication. #5442 covers all-width caller
algebra, while #5445 explicitly leaves actual control reservations open.

The earlier proposed generic four-update audit was withdrawn after finding
`compact-carry-repair-audit`, `compact-late-guard-audit-seed73`,
`compact-agreement-set-seed29`, and PR3055 `physical-complex-compact`.
`test_compact_control_layer.py` already checks generic toy row splits;
`physical-complex-controller` handles an older73-axis caller. We reuse their
boundaries and target this exact selected source, not a new generic repair.

## Open next questions

- Bind the unmaterialized final compiler's occurrence maps to this address ABI,
using actual N/W/O and full stock products. Start with one chronological
product occurrence and its rank-child embedding; the missing source-derived
map is the blocker, not another raw-route census.
- Join the actual selected packet to the encoded physical controller while
preserving descriptors, parent return and compact-field restoration. Begin
with one IDLE packet. This package proves address placement under those
inherited execution premises; it does not implement that join or price it.

## Credit

Prepared for **JoyboyBrian**, with substantial **OpenAI Codex assistance**.
Apache-2.0. Retain Douglas Colkitt's compact/row arguments; John-Michael Smith,
Hanson Sun and Aaron Wang's prior audits; Jose Manuel's physical implementation;
Chris R / Clear1nsight's PR5413 caller, Dillon's provider, Ali Romman's relations,
Cao Yifan's native word, Cinna Davis's controller, zxv's compiler and Jacob Alan
Sussman's recursion, plus the original source notices and cumulative lineage.
No independent human review, kernel proof, selected promotion or paid gain.
70 changes: 70 additions & 0 deletions research/selected-control-reservations/RESULT.json
Original file line number Diff line number Diff line change
@@ -0,0 +1,70 @@
{
"central_translations": 16,
"idle_child_cases": 45,
"literal_lift_coordinates": 207060,
"local_planes": 18,
"placement_families": [
{
"active_end": 60992,
"f": 1,
"origin": 6848,
"q0": 11,
"qB": 47,
"qF": 94
},
{
"active_end": 121984,
"f": 2,
"origin": 12864,
"q0": 11,
"qB": 94,
"qF": 188
},
{
"active_end": 182976,
"f": 3,
"origin": 18880,
"q0": 11,
"qB": 141,
"qF": 282
},
{
"active_end": 426944,
"f": 7,
"origin": 42880,
"q0": 11,
"qB": 329,
"qF": 657
},
{
"active_end": 1890944,
"f": 31,
"origin": 186880,
"q0": 11,
"qB": 1454,
"qF": 2907
},
{
"active_end": 61000000000000000000000000000000000,
"f": 1000000000000000000000000000000,
"origin": 6000000000000000000000000000000832,
"q0": 11,
"qB": 46875000000000000000000000000000,
"qF": 93750000000000000000000000000000
}
],
"routes": 119,
"row_boundary_cases": 21,
"scope": "raw5413 route/reservation binding; expanded compiler injection and stock unverified",
"semantic_controls": [
"IDLE_wrong_pre_direction",
"route_out_of_template",
"stage_private_axis_collision",
"omitted_central_translation",
"active_slot_in_front_control",
"missing_child_spectator",
"nondivisible_row_stock",
"undersized_active_region"
],
"swaps": 16748
}
54 changes: 54 additions & 0 deletions research/selected-control-reservations/SOURCES.json
Original file line number Diff line number Diff line change
@@ -0,0 +1,54 @@
{
"schema_version": 1,
"base_commit": "32890bc5de034ab8fc93ac88ead7279a6a416d04",
"files": [
{
"path": "certificates/selected-result.json",
"usage": "provenance",
"git_blob": "fc3d4ba105e1f097d0d7d051a308281e86ab1446",
"sha256": "66974efdced4083c00025f97d95d83909a226cdf801edb9301c195c6641f6903"
},
{
"path": "docs/research/agent-task-catalog.json",
"usage": "provenance",
"git_blob": "f126dd2e2dad2b669e75d1d0d0bd05c600725602",
"sha256": "f7a48bfa2659b88378282a8eee7d9fe52cdf505594124c4d1686b58510fffd72"
},
{
"path": "notes/compact-control-layout.tex",
"usage": "input",
"git_blob": "16f4d233471a7d189163ae7e7902becadecbba64",
"sha256": "2bc28e50f3eb092195294cafa81213e520f6d42e7935e9905ad8a6199e2a7c87"
},
{
"path": "notes/compact-control-movement.tex",
"usage": "input",
"git_blob": "06b6a9ef24245ac7b724c53a204635415b402357",
"sha256": "b963c3d2f4e1b7059aa289bb4b4ac89b688920bcf42f5b20d7698cdef0132cc6"
},
{
"path": "research/coupled-helper-neutral-completion/capsule/workspace/swarm-coupled-caller-sol-high-20261024/agents/challenge/WHOLECALLER-PLAN-v1.json",
"usage": "input",
"git_blob": "b868fe46ce7ff96a51f7b612758685a23014d31d",
"sha256": "a02e79fb4615f4f987e3b3617c2b42eea2c1d7415cc66fa702a8b0cd28e6387e"
},
{
"path": "research/coupled-helper-neutral-completion/reports/proofs/V16-APPLICATION-PROOF-v3.md",
"usage": "input",
"git_blob": "574c1d0bd75cd85f720991d815272738aa3e5b8f",
"sha256": "cda8f596f7854f42a228f757d5f8e68f6b2256a19c5ae3771fb40753cedc5b39"
},
{
"path": "research/physical-complex-controller/PROOF.md",
"usage": "input",
"git_blob": "c505caecc7ced19a4efc35c423efedda9427bea5",
"sha256": "471185621c0ecce295ffd327b1244aed234efe5ff05a804aaaa11316f1a8319e"
},
{
"path": "tests/test_compact_control_layer.py",
"usage": "input",
"git_blob": "bca34c9e796ce5483f99ae1bf31acc6b74d32d8c",
"sha256": "c39797d9a2bc7eb4b9c7ffc458be04b64806636a3ef544de3a273f11ebfbcea1"
}
]
}
Loading