Complete five-speed bound and tight equality classification

This checkpoint verifies an unconditional lower bound at 1/6 for at most five distinct positive integer speeds and the complete primitive and gcd-normalized five-speed classification at that threshold. It is an independent finite proof of the bound already registered through the Renault formalization. The full five-speed spectrum below 1/5 remains separate.

Exact statements

Loneliness.Modular.FiveRunner.five_runner_bound and Loneliness.CoordinateSubspace.five_speed_lonelyRunnerBound prove CoordinateSubspace.LonelyRunnerBound 5: every nonempty list of at most five distinct positive natural-number speeds has actual loneliness at least 1/6. No smaller-runner, geometric, prime-classification, or finite-certificate hypothesis remains.

For every positive v : Fin 5 → ℕ with gcd one and actual loneliness at most 1/6, Loneliness.Modular.FiveRunner.primitive_classification proves that List.ofFn v is a permutation of exactly one of these retained directions:

  • [1, 2, 3, 4, 5];
  • [1, 3, 4, 5, 9].

Both retained directions have actual value 1/6, and primitive_value_eq proves equality for every tuple satisfying those hypotheses. The primitive classification does not require distinct coordinates as an additional premise.

complete_normalized_classification applies to every positive list of length five with actual loneliness at most 1/6. Its gcd-normalized list is a permutation of one of the same two literal directions. This supplies the complete tight-five classification; it does not classify other values in the interval (1/6, 1/5) or complete the five-speed spectrum theorem.

The generated audit source contains four explicit quantified assertions with these literal lists. The complete named obligation inventory is preserved in receipt.json.

Finite proof and certificate coverage

The geometric argument proves squared norm at most 347142 for a positive primitive five-speed vector with value at most 1/6. The forced divisor 60 and the exact prime-product margin yield a prime dividing none of its five coordinates. The selected primes are 149, 151, 163, 167, 173.

Prime Minimum-pair roots Support frames Retained rows
149 73 1171 10
151 74 1478 10
163 80 1596 10
167 82 1430 10
173 85 1395 10
Total 394 7070 50

Every support of size at most five is covered. The 118 generated prime modules reproduce byte-for-byte from the published producers. Their frame proofs state the original bounded-check propositions. The Fuel checker has proved unconditional equality with the original checker, and ordinary decide +kernel checks the numerical propositions. The source generator and search code do not become trusted mathematical assumptions.

Verification evidence

GitHub Actions run 44549751944, job 135649983909, completed successfully at 1ffeb0932ac9a5674f2d416757d786a85e43ffde. It compiled all 250 project sources afresh with Lean 4.35.0, one worker, -M4500, and warnings treated as errors. Fresh source checks took a total of 2324.416 seconds; the final theorem source took 2.757 seconds.

The audit checked 28,383 owned/raw declarations, including 23,357 theorems and 118 private declarations. All 7,586 required theorem endpoints and all four literal statements passed. The only permitted transitive axioms were propext, Classical.choice, and Quot.sound. The unchanged stock leanchecker then replayed exactly the 250 staged project modules, with no missing, repeated, or extra module, in 1804.712 seconds.

Artifact 14551506873 has ZIP SHA-256 018afd2acad8e32dff6ea07429e068afc8894e43f8375becb4df13e38d76e6c1. The unchanged original receipt has SHA-256 4bcd923b04ce812cab68abd91ed8c6ee102f594a1f59bd310752dbfd79aeb542.

Authentication checked all 1,009 archive members. It matched all 250 proof sources and 21 recorded controls to their original Git blobs, pinned all 17 imported repository Python modules, reproduced every finite certificate, and compared every source/audit/replay process record and materialized object facet with its recorded hash. It revalidated the complete raw/private axiom inventory and exact replayed-module set. The generation inventory matches byte-for-byte; the generation log matches after excluding only elapsed-time measurements. See artifact-authentication.json and published-inputs.json.

Preserved bytes and reproduction

All 509 original receipt, generated-audit, generation, log, and process-record files are preserved. Only the two largest files are compressed: declaration-inventory.json.gz and logs/all-owned.audit.log.gz. Decompression restores the exact original bytes. preservation.json records every original and stored path, encoding, byte length, and SHA-256. Project proof sources remain available at the recorded Git commit; object facet hashes remain in the original receipt and the downloadable artifact.

At the recorded commit, use the pinned toolchain and dependency revisions:

python3 scripts/verify_five_runner_exact.py --plan
cd lean
lake env python3 ../scripts/verify_five_runner_exact.py

This invokes the fresh gate. Its imported Mathlib and toolchain packages are pinned dependencies; the receipt does not claim that those packages were rebuilt from source. Registration and the subsequent merged-package gate remain distinct from this original complete component verification.