[Clear1nsight]

Change and mathematical scope

Certify a single-point optimistic decoder allowance for the frozen Global120 no-donation model. At C*=.00330007, rho=10^-9, and tariff r^(1-C*), the independent exact interval calculation gives

916.028913884418026624304657 <= (1-rho)T(C*)-P(C*) <= 916.028913884418026624304658.

For additional paid moment A and lost-target moment L, the necessary fixed-profile gate is A+(1-rho)L < B. The isolated additional-paid maxima are 916 H1, 115 H8, 57 H16, 38 H24, 37 H25 or 6 H145. These alternatives spend the same allowance and are not additive. Inventory-capped target-loss counts are 0/115/0/0/37/6 at those ranks.

All 256 proposed new mode outputs are unearned, old-main preservation is an optimistic grant, and future decoder payments and target losses remain UNKNOWN. Their empty price profiles are an optimistic relaxation. This is no new physical caller, funded κ, higher-moment/tail/compiler/precision/stock result or outer admission. No claim extends this pointwise screen to larger C or arbitrary circuits.

The exact actual-neutral baseline is 64,260 paid /10,928 targets, not the historical 64,404-call arithmetic profile. The declared successor is 64,668 paid /10,928 optimistic targets, with rank masses248,598/250,366. Its complete signed delta charges both endpoint pads and debits 256 distinct reused H8 outputs exactly once. The 1,768 rank-mass surplus does not itself establish the tested moment or a construction.

Task and overlap

This follows the donor's assigned independent cost/verification support for the ongoing κ≥.0033 research direction. The actual physical decoder, original construction and other separately owned fanout work remain separately owned; this contribution performs no replacement census or constructor search. The package declares the scope change against its pinned task.

Canonical provenance is 1a3ced60b2f7a46bc70280a04f2eeb8ca027c700, selected conditional κ .003027288235 under the same five assumptions. The proposal, its independent 396-check intake and the new numerical receipt are explicitly LOCAL generated evidence with permission to vendor their unchanged bytes; they are not claimed as canonical ancestors. Retained manifests and source-specific notices bind the exact inputs.

PR #5423's eight-H1 component saving is already published and is not silently subtracted from this Global120 invoice. Its standalone28-H1 gate, and #5418/#5420's older fixed-model allowances, do not govern this changed profile. PR #5413 supplies the accepted neutral baseline. The sole all-author PR feed was refreshed through 07:43:15 UTC; canonical main/guidance and relevant PR heads were unchanged and no competing budget publication appeared in that bounded stream. The newest #1166 handoff was also read in Browser and remained the existing #5423 component. These checks are bounded, not exhaustive novelty certification.

Validation

The sole peer's original guarded execution used the unchanged checker ba5b1a7d1e1c9de6a485e1e75947fa37c370d3a1f7d7dc0c5ccfa309e65e6b6e and produced complete raw receipt 406b81580297e81b13079c81afa28796f3f41c947586de164eaef8e721e1e659. It exited0 without a resource stop in 0.303001 seconds, sampled peak 40,431,616 bytes. The same cached powers serve rho0 and rho1e-9; no second C, root search or historical scientific suite ran.

The independent static audit authenticates 15 source and 3 assessor bindings, all 40 pre/post/current byte-size-mode records, grant/actual argv/guard/stdout agreement, rational endpoint consistency, 23 outward decimal fields and strict quota/next-integer boundaries. It computed no fresh powers and ran no submitted scientific engine. The source-independent 396-check intake is retained provenance and is not rerun by the hook.

The frozen package of 48 files passed the sole peer's three serial 30-second/256-MiB requests: registered normal/optimized pair 1.813470s,110,690,304 sampled peak bytes; readonly normal .605000s,68,849,664B; readonly optimized .806149s,82,268,160B. These are four fresh executions of the same fixed-point checker, separate from the original screen. All four full scientific receipts match typed canonical JSON projection 7f6598cbeec4f1ec73916056fb675aa4dd3698318b8c9d7342d22023258e59af. Only environment.elapsed_seconds and environment.maximum_resident_set_kib are excluded after exact-key/finite/nonnegative validation; every other field, numeric type, fraction, target flag and unknown/null obligation is compared exactly. Raw full receipts are retained.

Ten wrapper controls reject forged provenance, altered raw/numerical evidence, unearned-target or zero-cost promotion, additive quotas, invalid environment fields/types and scientific Boolean/integer or float/integer substitutions. The two numeric-type controls augment the original eight controls still described by the frozen VERIFY prose. The initial comparison wrapper was rejected statically before any execution; its scientific checker and accepted source result were unchanged. No failed scientific run was repeated.

All 101 pre/post/current file byte-size-mode records (48 working, 48 readonly, 5 tool/request records) and eight readonly directories match. Portable files were 0444/directories 0555 with empty PATH, absolute Python, explicit Git-discovery ceiling and unset Git worktree variables. Ancestor Git metadata was preserved; Git discovery under the ceiling returned 128. An outer metadata-helper SyntaxWarning is preserved in the raw execution record; the guarded wrapper/checker runs and child stderr show no scientific warning or failure.

The passive public receipt certificates/research/global120-optimistic-budget-package-validation-v2.json, SHA256 073b976ac7bbae6f50defb183b740e84bb7b6dc4ac46ccb5bfd17aaef6824e06, retains all four raw fresh receipts, three guards/logs and full execution custody. It is outside the 48 executed package sources and source export, and is not claimed as an executed input or a selected-result certificate. Frozen README/request preparation-status text is historical; these actual receipts supersede its pre-execution pending status.

Science/evidence commit: 280690fdf35b4d944dfcc1bc54c3dc13c15619dd, direct parent canonical 1a3. Publication head: e638ed60241c7a04e49c79ab1f98d612a94e0c11. The 49 science/evidence paths comprise 48 executed package sources plus the passive actual-validation JSON; the publication has 50 paths after its sole extra CI documentation file. All 48 package payloads match the executed working and readonly sources, both commits and official export. Metadata, committed-byte and declared-source-history checks pass. The canonical commit graph contains 7507 commits; a complete file-corpus checkout or historical mathematical replay is not claimed.

Official export/check passed: SHA256 272ce51218a94d08af2e6ea2d759d9ddd79c7c18226dba68d865dcece9314da6, 11,028,480 bytes, 49 members comprising 48 payloads plus one transport receipt. Its payloads are byte-bridged to already executed source; the final archive was not separately rerun. Working 0644, readonly 0444/directories 0555, Git 100644 and export 0644 are distinct representation modes. The passive validation JSON is public separately and is not in this source export. Registered package; complete passive execution receipt. Publication also carries the exact previously tested #5420/#5423 docs/ci-verification.md navigation repair, because it is still absent from canonical main. No mathematical source is changed by that separate documentation commit.

python3 -B scripts/research_package.py --package research/global120-optimistic-budget
python3 -B scripts/check_research_commit.py --package research/global120-optimistic-budget --check-source-history

The hook executes only the independent fixed-point interval screen and its wrapper controls, not the primary constructor, original source assessor, native/phase engine, selected caller, compiler or Lean. Direct portable command is /usr/bin/python3 -I -Werror -B verify.py, adding -O for optimized mode. The metadata/history command executes no scientific checker.

First exact-head hosted workflow 40075828407, attempt 1, completed SUCCESS at e638ed60241c7a04e49c79ab1f98d612a94e0c11. Plan/onboarding/selection and the registered bounded replay with committed inputs unchanged passed. Actual CI logs show fresh normal and optimized interval-screen results with the same typed scientific projection and all ten wrapper controls. This hosted pair is separate from the four local registered/portable executions. It retains all 256 modes unearned, optimistic old-main preservation, unknown future costs, no complete caller and funded_kappa=null; hosted success is no new κ or scientific admission.

Credit

Peer Global120/chart/invoice authors; Ali Romman/controller and cached-profile lineage; Dillon/provider; Cao Yifan/native and Clifford work; Cinna Davis, zxv, Jacob Alan Sussman, Dugongue, Chris R/Clear1nsight and retained native/helper/BIT/compiler/precision/analytic predecessors. Credit the independent source assessor and one-point numerical/verification lanes. Substantial OpenAI Codex assistance; Apache-2.0 and inherited source notices retained.

Open next questions

1. Can an actual all-input, version-correct Gamma decoder and return path fit the complete mixed allowance? First supply its literal chronological word, raw inverse transpose, owner/restoration proof and additional-paid/lost-target ledger.
2. Can the 256 proposed mode outputs and old-main outputs be realized as independent canonical Gaussian jobs? Bind each original input, all phase/chart conversions and final normalizers once; optimistic reward is not target evidence.
3. Can the completed caller fund the actual higher moments, tails, finite compiler/reserve, ordinary routes, precision, simultaneous stock and outer transfer for κ≥.0033? Those require fresh source-bound applicability and independent validation after the physical caller exists.