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
28 changes: 28 additions & 0 deletions research/staged-width-t488/CLAIM.json
Original file line number Diff line number Diff line change
@@ -0,0 +1,28 @@
{
"replicas": 488,
"calls": 4392,
"roles": 1791936,
"physical": 437757,
"profile": {
"1": 10826280,
"2": 3188592,
"3": 1015650,
"4": 447984,
"5": 94794,
"8": 35136,
"48": 186782
},
"root_interval": [
".001730096363528623534230",
".001730096363528623534231"
],
"candidate_endpoint": "17300963635286235342299999999999999999999998999999999999999999999999/10000000000000000000000000000000000000000000000000000000000000000000000",
"frontier_status": "Published PR3626 complete source-bound conditional release; both hosted checks passed",
"frontier_sha256": "9f27717c3ddbf5eb7b9773e1b2887ff3899d354019139d916c9b8004aee2e097",
"scope": "Complete source-bound conditional release with authenticated same-session native reuse; inherited mathematical premises remain.",
"frontier_pr": 3626,
"frontier_head": "659fd5e7064d9466b668d7188cbb013b333a415a",
"status": "PASS_COMPLETE_CONDITIONAL_RELEASE",
"kappa": "17300963635286235342299999999999999999999998999999999999999999999999/10000000000000000000000000000000000000000000000000000000000000000000000",
"evidence": "evidence/EVIDENCE-PINS.json"
}
9 changes: 9 additions & 0 deletions research/staged-width-t488/CREDITS.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,9 @@
# Credits and provenance

Ross Richardson’s PR3605 supplies the staged-width operators, T37 lineage and integral-network allocation machinery. Our published T530 contribution PR3626 is the comparison baseline. This candidate reuses both the retained T8 chronology, also used in our T450 contribution PR3624, and the retained T16 chronology used in T530.

The retained T16 chronology originates in PR3413 and is preserved in our PR3457/3465 provider-composition lineage. T16-PROVENANCE.json pins the actual PLAN and reconstruction source. This contribution supplies the combined chronology and newly solved allocation; it does not claim the constituent chronologies or count their prior gains twice.

Retained machinery preserves its upstream contributor notices and exact sources, including the earlier rank-seven compiler/controller and conditional-release chain. Private optimizer tools are excluded. New material is Apache-2.0, prepared for hexapode with substantial OpenAI Codex assistance. No independent human or formal-kernel acceptance is claimed.

T488 adds the three-stream chronology and joint allocation, using both exact retained T8 and T16 orders. Historical supplier modifications remain attributed upstream. The ceiling proof is a fixed-model allocation result, not a new operator.
128 changes: 128 additions & 0 deletions research/staged-width-t488/EXECUTION-COVERAGE.json
Original file line number Diff line number Diff line change
@@ -0,0 +1,128 @@
{
"status": "PASS_TERMINAL_T488_FULL_ADMISSION",
"authoritative_session": 48234,
"exit_code": 0,
"source_pins": {
"/private/tmp/integer-mult-experiments/contract/pr3605-t488-full-admission/RESULT.json": "50698ed610dd4b2ebfab2c9a92b657f16d3e1987da8d564c13a588a25b1437e4",
"/private/tmp/integer-mult-experiments/contract/pr3605-t488-full-admission/release/RELEASE-RECEIPT.json": "dd5c6eea88986c57ccd9ce76c2702c60817237db1935729568134dbd001007b5",
"/private/tmp/integer-mult-experiments/contract/pr3605-t488-full-admission/CURRENT-FRONTIER-CHECK.json": "0abf49110acaea4e176abcdc2ccf667852cdcb9137115e14bed8a26c13854861",
"/private/tmp/integer-mult-experiments/contract/pr3605-t488-launch.json": "695d1fb0f4c74d5e49d67af05727158b53b972a8ec8f211e5ec50c0d15e8e1ce",
"/private/tmp/integer-mult-experiments/contract/pr3605_t488_entry.py": "00cb1277497c053c7ca2b7d58c2a95a14eac22841b3f203f90e5533e3cfd30d3",
"/private/tmp/integer-mult-experiments/contract/pr3605-public-full-replay/changed/native/NATIVE-RECEIPT.json": "411b6fc1ac480263a90d36eca58c109687307e9a2d204af3ed4d002f9f8ff629",
"/private/tmp/integer-mult-experiments/contract/pr3605-t488-portable-live-equivalence.json": "835e929bf358077efdc3461a02946e19377fff19a1aab3789aeb80b070dad8ab"
},
"stages": [
{
"name": "owners",
"exit_code": 0,
"seconds": 32.97771345899673,
"command": [
"/private/tmp/integer-mult-experiments/adversarial/cap740-venv/bin/python",
"-B",
"/private/tmp/integer-mult-experiments/contract/pr3605-t488-full-admission/package/reconstruct.py",
"--out",
"/private/tmp/integer-mult-experiments/contract/pr3605-t488-full-admission/owners"
]
},
{
"name": "parent-lift",
"exit_code": 0,
"seconds": 67.28546375001315,
"command": [
"/private/tmp/integer-mult-experiments/adversarial/cap740-venv/bin/python",
"-B",
"/private/tmp/integer-mult-experiments/contract/pr3605-t488-full-admission/package/lift_parent.py",
"--package",
"/private/tmp/integer-mult-experiments/contract/pr3605-t488-full-admission/package",
"--owners",
"/private/tmp/integer-mult-experiments/contract/pr3605-t488-full-admission/owners",
"--out",
"/private/tmp/integer-mult-experiments/contract/pr3605-t488-full-admission/parent-data"
]
},
{
"name": "rebind",
"exit_code": 0,
"seconds": 6.395641707989853,
"command": [
"/private/tmp/integer-mult-experiments/adversarial/cap740-venv/bin/python",
"-B",
"/private/tmp/integer-mult-experiments/contract/pr3605-t488-full-admission/package/new/rebind.py",
"/private/tmp/integer-mult-experiments/contract/pr3605-t488-full-admission/package",
"/private/tmp/integer-mult-experiments/contract/pr3605-public-full-replay/changed/native/native-cache/735c540b1fe29d0ceeb4d2f304ec941e5f2bd3bc92bafc8cce2b4f9a67a3b653/local",
"/private/tmp/integer-mult-experiments/contract/pr3605-public-full-replay/changed/native/native-cache/886338fc32da14f3389a82c8120a1bcf9c5e2f0ec3e9b2d377fdae809e1d977f/local",
"/private/tmp/integer-mult-experiments/contract/pr3605-t488-full-admission/rebound"
]
},
{
"name": "discovery",
"exit_code": 0,
"seconds": 588.8035517499957,
"command": [
"/private/tmp/integer-mult-experiments/adversarial/cap740-venv/bin/python",
"-B",
"/private/tmp/integer-mult-experiments/contract/pr3605-t488-full-admission/layout/packing74/replay.py",
"--data",
"/private/tmp/integer-mult-experiments/contract/pr3605-t488-full-admission/layout/packing74/candidate",
"--work",
"/private/tmp/integer-mult-experiments/contract/pr3605-t488-full-admission/discovery",
"--local",
"/private/tmp/integer-mult-experiments/contract/pr3605-public-full-replay/changed/native/native-cache/735c540b1fe29d0ceeb4d2f304ec941e5f2bd3bc92bafc8cce2b4f9a67a3b653/local",
"--discover"
]
},
{
"name": "outer-check_ledger.py",
"exit_code": 0,
"seconds": 0.08240837501944043,
"command": [
"/private/tmp/integer-mult-experiments/adversarial/cap740-venv/bin/python",
"-B",
"/private/tmp/integer-mult-experiments/contract/pr3605-t488-full-admission/package/outer/check_ledger.py"
]
},
{
"name": "outer-independent_check.py",
"exit_code": 0,
"seconds": 0.03534187498735264,
"command": [
"/private/tmp/integer-mult-experiments/adversarial/cap740-venv/bin/python",
"-B",
"/private/tmp/integer-mult-experiments/contract/pr3605-t488-full-admission/package/outer/independent_check.py"
]
},
{
"name": "release",
"exit_code": 0,
"seconds": 891.9768864999933,
"command": [
"/private/tmp/integer-mult-experiments/adversarial/cap740-venv/bin/python",
"-B",
"/private/tmp/integer-mult-experiments/contract/pr3605-t488-full-admission/package/verify_release.py",
"--main",
"/private/tmp/integer-mult-experiments/contract",
"--work",
"/private/tmp/integer-mult-experiments/contract/pr3605-t488-full-admission/release",
"--native-cache",
"/private/tmp/integer-mult-experiments/contract/pr3605-public-full-replay/changed/native/native-cache",
"--jobs",
"2"
]
}
],
"kappa": "17300963635286235342299999999999999999999998999999999999999999999999/10000000000000000000000000000000000000000000000000000000000000000000000",
"profile": {
"1": 10826280,
"2": 3188592,
"3": 1015650,
"4": 447984,
"48": 186782,
"5": 94794,
"8": 35136
},
"native_coverage": "Fresh both-current-provider five-stage cache authenticated and reused; retained parent evidence separately authenticated. T488 did not rerun cold native providers.",
"portable_coverage": "Exact portable/live source and driver equivalence; no completed T488 portable CLI claimed.",
"admission_scope": "Conditional inherited all-size/compiler/precision assumptions remain. Internal supplier endpoint differs from global outer headline.",
"scientific_source_mutations_during_run": false,
"preflight_note": "One entry-only missing candidate filename rejected before scientific driver launch; corrected exact candidate path and preserved preflight failure log. No scientific rerun."
}
25 changes: 25 additions & 0 deletions research/staged-width-t488/EXPECTED.json
Original file line number Diff line number Diff line change
@@ -0,0 +1,25 @@
{
"replicas": 488,
"calls": 4392,
"roles": 1791936,
"physical": 437757,
"profile": {
"1": 10826280,
"2": 3188592,
"3": 1015650,
"4": 447984,
"5": 94794,
"8": 35136,
"48": 186782
},
"root_interval": [
".001730096363528623534230",
".001730096363528623534231"
],
"candidate_endpoint": "17300963635286235342299999999999999999999998999999999999999999999999/10000000000000000000000000000000000000000000000000000000000000000000000",
"frontier_status": "Published PR3626 complete source-bound conditional release; both hosted checks passed",
"frontier_sha256": "9f27717c3ddbf5eb7b9773e1b2887ff3899d354019139d916c9b8004aee2e097",
"scope": "Finite profile expectation; no completed T488 admission.",
"frontier_pr": 3626,
"frontier_head": "659fd5e7064d9466b668d7188cbb013b333a415a"
}
27 changes: 27 additions & 0 deletions research/staged-width-t488/FRONTIER.json
Original file line number Diff line number Diff line change
@@ -0,0 +1,27 @@
{
"replicas": 530,
"calls": 4770,
"roles": 1946160,
"physical": 475389,
"profile": {
"1": 11757460,
"2": 3463020,
"3": 1102502,
"4": 486540,
"5": 103690,
"8": 38160,
"48": 202828
},
"root_interval": [
"0.001730071864705779882540",
"0.001730071864705779882541"
],
"candidate_endpoint": "17300718647057798825399999999999999999999998999999999999999999999999/10000000000000000000000000000000000000000000000000000000000000000000000",
"frontier_pr": 3624,
"frontier_head": "515a7c4490aa32ac3ca5014feb6f7da0d6952863",
"frontier_sha256": "8f2f80438593e7efaa6bb57c881a40cedbd7b5665b4ae12c6cf215d217c98397",
"scope": "Complete source-bound conditional release with authenticated same-session native reuse; inherited mathematical premises remain.",
"status": "PASS_COMPLETE_CONDITIONAL_RELEASE",
"kappa": "17300718647057798825399999999999999999999998999999999999999999999999/10000000000000000000000000000000000000000000000000000000000000000000000",
"evidence": "evidence/EVIDENCE-PINS.json"
}
11 changes: 11 additions & 0 deletions research/staged-width-t488/PROOF.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,11 @@
# Scope and finite argument

The three source orders and dispatches are pinned independently. Their concatenation is re-executed against the retained carrier template to derive every open/closed owner window and physical family. The actual changed-provider widths and paid local profile are authenticated. Aggregate integer checks enforce every call demand, group-width balance, bin count and shared group/time degree.

Each pattern has at most five helper slots and S=8 copies. The retained integral copy-coloring construction decomposes macrotime tasks into eight actual calls, then assigns bins to copies using required endpoint copies. Macrotime flow is not a single call: the earlier one-role-per-bin-per-macrotime inequality is not imposed. A separate literal auditor checks all rows against source-derived windows and actual native frames.

Let F(k)=sum_r n_r (r/73)^(1-k)/(896T). All paid ranks are below73, so F is strictly increasing. Total rank mass is65088T, below65408T=73*896T; the rational interval checker brackets the unique crossing. Comparison of different T values uses530*n488−488*n530, not raw profile subtraction. The candidate endpoint subtracts1e-46 and1e-70 from the lower root bound. This algebra is a finite cost statement; it does not itself establish the complete algorithm.

The separate ceiling certificate checks all2551 nonnegative capacity patterns, exact dual inequalities and equality for the candidate. Its canonical projection includes source order/dispatch, groups, widths, inventory, local profile, x/z and paid profile; private discovery metadata is omitted without changing these fields.

The complete source-bound release passes supplier, stock, prefix, precision, physical and outer checks. Its authenticated invoice charges physical stock437757, stock logarithmic bound1324, prefix1479 and reserve1512. Stored receipts are checked separately from a fresh full replay. No stored T450/T530 owner or release receipt is substituted for a T488 replay.
34 changes: 34 additions & 0 deletions research/staged-width-t488/README.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,34 @@
# T488 three-chronology allocation (completed conditional admission)

This candidate combines the retained T37, T8 and T16 source orders with the unchanged PR3605 staged providers. It recomputes their joint windows and supplies a new integer allocation. It does not add the older headline gains or claim a new local operator.

All 1,791,936 literal roles passed an independent native-width, window, family, offset and invoice audit. The exact normalized root lies between 0.001730096363528623534230 and 0.001730096363528623534231. All seven source-bound admission stages passed. The conditional endpoint is `0.001730096363528623534230 − 10^-46 − 10^-70`. The T530 comparison is now a completed source-bound conditional release; its publication is PR3626 at659fd5e7064d9466b668d7188cbb013b333a415a (both hosted checks passed).

From the repository root, run `python3 -B research/staged-width-t488/verify.py` for finite source, aggregate, pricing, ceiling and rejection controls. The owner builder is standard-library deterministic reconstruction, with no optimizer dependency. See owners/README.txt for literal reproduction. Full replay is a separate command through replay.py; inherited all-size/compiler premises remain.

The ceiling certificate covers all packing patterns for the fixed supplier widths, capacities and nine-call normalization. It does not constrain changed suppliers, geometry, recursive interfaces or physical contracts.

## Open next questions

Further fixed-inventory repacking cannot improve the certified relaxed optimum; stronger results require changing its stated model.

From the repository root, reconstruct the literal allocation into a fresh directory:

```sh
OWNER_SOURCE_REPO="$PWD" python3 -B research/staged-width-t488/owners/portable_expand.py --out /tmp/t488-owners
```

Then run complete replay into another fresh directory:

```sh
python3 -B research/staged-width-t488/replay.py --work /tmp/t488-release --allocation /tmp/t488-owners --main-files . --t8-package research/t8-rank7-two-swaps
```
This may take substantially longer and is not part of the finite registered check. The package authenticates five stored terminal receipts separately from fresh replay. The completed execution reused authenticated native caches freshly checked earlier in this session. Its source manifest and scientific driver match the portable materialization, but the complete fresh portable CLI has not run end to end.

## Review corrections (PR3694)

Sagar Thakrar's independent [PR3694](https://github.com/CrocSwap/integer-mult-bounds/pull/3694) reproduced the literal allocation, invoice and fixed-inventory ceiling, and identified two inherited checker defects. The owner auditor now routes both real data and corrupted fixtures through the same five validation rules. The finite replay includes independent production-callsite tests: disabling each rule, all five rules, or the production call itself is detected. These small fixtures exercise the actual extracted production statements; they are separate from the complete literal replay.

Fresh native canonical authentication now hashes decompressed bytes, accepting gzip header or compression differences while rejecting changed payloads. Historical container hashes and the frozen admission receipts remain unchanged. Portable tests exercise the actual authentication loop using synthetic payloads; local checks also covered both retained native providers.

The corrected auditor passed a local replay of all 1,791,936 literal roles, with the same profile and physical stock. This correction does not change κ or the scientific program, and is not a new full native or seven-stage admission run. The registered finite checks include both regression tests.
Loading