[1337isnot1337]

PR #4940 leaves the finite compiler constants for its declared coded p19 inventory open. This package supplies a strict arithmetic instance with 128 chronological copies and logarithm power 140, replacing the failing inherited 1024/1000 choice.

The independently reconstructed inventory retains all 19 complete native/helper profiles, 20 thin queries and both IDLE tails: 12,008 targets and 68,782 paid calls. Two separate exact rational interval implementations certify both completion/splitting tails and shifted compiler capacity. Exact integer arithmetic gives the sufficient normalized compiler rank encoding 2^4106900000000. All 54 strict outer predicates and four functional guards pass at the planning endpoint 0.00339999, conditional on realizing this declared inventory under the source physical and compiler contracts.

No new admitted κ. This does not establish a source-bound physical p19 caller or identify its coded outputs as independent Gaussian jobs. Physical code/frame mixing, odd-denominator recoding, arbitrary-input/helper returns, complete nonrecursive precision/storage/tape work and the inherited BIT/analytic/recovery premises remain explicit prerequisites. The two active physical-pair and denominator-13 claims in Discussion #4930 remain distinct work; further compiler parameter scans do not resolve them. The selected record is unchanged.

Validation: registered Python 3.12 replay; optimized direct replay; metadata/source pins; committed-byte preflight; official 17-file export and integrity check; optimized export replay with no Git on PATH and networking disabled. Controls reject omitted query charges, the inherited compiler choice, insufficient conservative tensor power, closed outer gaps and insufficient BIT headroom. The optional full source-history check is unavailable because this checkout is shallow; this is recorded separately from the passing committed-byte check.

Canonical claim: #4930 (comment). Written derivation and scope are in research/coded-compiler-closure/PROOF.md; exact receipt and every remaining prerequisite are in RESULT.json. Open next questions and predecessor credit are in the README and task metadata.

Credit SethBurkart123's PR #4940; Dugongue/mdumitrean's PR #4929; Dillon's PR #4933; Cinna Davis's PR #4440/4470; Cao Yifan/intqwq's PR #4610; zxv's higher-moment/compiler and interval work; Jacob Alan Sussman's mixed flow; and retained native, Clifford, BIT and outer contributors. Prepared by Cinna Davis with substantial OpenAI Codex assistance. Apache-2.0; no independent human or proof-kernel acceptance claimed.

[1337isnot1337]

Upstream checkpoint at 4c0a359d88f22f05189d51f48387750b801a45d4: the newly archived physical-pair counterexample proves that PR4940's literal scalar inverse-code substitution leaves the nonzero residual D(I-A^-1 C). The actual physical code C=K^-1(A tensor I)K differs from scalar A; the declared X-only boundary recodings do not repair that block.

This is now a concrete failed premise for applying this PR's conditional compiler arithmetic to that literal recipe. The 128/140 compiler inequalities and planning endpoint remain statements under the explicitly declared inventory/physical hypotheses; no supplier, improved K or complete caller follows. A different source-bound physical continuation and complete invoice are required. The archived counterexample does not prove impossibility for every later continuation.

[Matarener]

Note: the coded p19 compiler/outer arithmetic after the new ordinary BIT (#4760/#4814). No new κ.

#5027 certifies the p19 inventory at the planning endpoint 0.00339999 against the old BIT literal 17/5000. #4814 (merged) now admits ordinary BIT 1737729/500000000 = 0.003475458, so I re-ran the same arithmetic with only bit_saving replaced and the planning endpoint moved:

• Outer ledger (54 strict + 4 functional checks; Consume the #4929 center-free word with coded reverse stages and A⁻¹-mixed query shears (no κ) #4940/Certify coded p19 compiler reserve at a conditional planning endpoint (no new κ) #5027 harness): with the new BIT value, the only binding predicate is still actual_bit_above_effective, so κ < 0.003475458 − 4·gap. That is 0.0034754551 at Consume the #4929 center-free word with coded reverse stages and A⁻¹-mixed query shears (no κ) #4940's gap 1/1.4e9 and 0.0034754523 at Certify coded p19 compiler reserve at a conditional planning endpoint (no new κ) #5027's gap 1/7e8. All the other 57 checks pass on the whole interval. Control: with 17/5000, 0.0034 fails only that predicate, as in Consume the #4929 center-free word with coded reverse stages and A⁻¹-mixed query shears (no κ) #4940.
• Consume the #4929 center-free word with coded reverse stages and A⁻¹-mixed query shears (no κ) #4940 p19 tail guards at C = 0.0034754551: α = 0.99587, γ = 0.9638, δ = 0.9516. They pass, and p = 19 is still the only p in 4..24 that passes all three.
• Certify coded p19 compiler reserve at a conditional planning endpoint (no new κ) #5027 compiler reserve at PLAN = 0.003475452 (TESTED = PLAN + 1e-8), with 128 copies, θ = 63/64 and the same interval engines:
  • Logarithm power 140 (and 142) fails the η > θ^power test. 143 passes and 142 fails. The new height is 2^4197900000000 (was 2^4106900000000).
  • The tails, the independent enclosures, the 54+4 ledger and the BIT-cap control (κ = new BIT is rejected) all pass. PLAN = 0.003475453 fails the ledger.
  • Control: the unmodified endpoint 0.00339999 with power 140 reproduces Certify coded p19 compiler reserve at a conditional planning endpoint (no new κ) #5027's height exactly.

So if the physical coded caller is admitted, the conditional planning ceiling of this line moves from ≈0.0034 to ≈0.0034755 (+2.2%), at the price of logarithm power 143. Separately, #4814's admission chain loses only ≈6e-11 between its raw root and its admitted value, so the C0/C1 tolerance question left open in #4697 is settled by #4814.

The scripts are copies of research/coded-reverse-controller/verify.py §5 and research/coded-compiler-closure/verify.py, changed only in bit_saving, the endpoint and the power. I can turn this into a package if useful.