This directory contains configuration files for verifying the proofs in Cryptids/
and BusyBeaver/ against the self-contained challenge specifications in
Challenge.lean and BusyBeaverChallenge.lean
using leanprover/comparator.
Challenge.leanimports only Mathlib and Lean core (Lean.Elab.Term), defines all conjecture-checking baseline and candidate machines (referenceandcandidate), states the target mathematical propositions (GoldbachConjecture,MatiyasevichRiemannCounterexample,ZFProvable,WadeZFProvable,PAProvable,paHalts,paZFInconsistent,ErdosTernaryConjecture, etc.), and states all 36 target state-reduction, equivalence,PAmetatheory & internalizedZFprovability, and semantics theorems withsorry.BusyBeaverChallenge.leanimports only Mathlib and Lean core (Lean.Elab.Term), defines the sharedBusyBeaverPost-Turing execution model, acceleratedTapePatternSet,StackPatternSet, stack/odometer/transducer reductions, and Collatz-like acceleration semantics (IsClosedPatternSet,HasCollatzReduction,IsClosedStackPatternSet,HasStackedCollatzReduction,HasStackReduction,HasOdometerReduction,HasTransducerReduction,IsValidAcceleration,CollatzHalts,IsHalt,IsNotHalt), andload_rle_machine%once, and states all 735BB(6)andBB(3, 3)holdout theorems (is_closed_pattern_set/is_closed_stack_pattern_set,has_collatz_reduction/has_stacked_collatz_reduction,has_stack_reduction,has_odometer_reduction,has_transducer_reduction,is_closed_segmented_system,is_valid_acceleration,isHalt_iff_acceleration,is_valid_highLevel_acceleration,isHalt_iff_highLevel_acceleration,acceleration_iff_highLevel_acceleration, A6is_not_halt,isHalt_HHH1LC_iff_HHH0LA, the 74HardnessEvidencelive-hazard witness and halting residue-class theorems across the 95 machines, and the Hard holdout theorems) withsorry.
Comparator exports both the challenge module (Challenge or BusyBeaverChallenge) and the
solution module (Cryptids or BusyBeaver) via leanexport, verifying in a Bubblewrap sandbox
that:
- Every target theorem and all of its transitive definition dependencies match identically between the challenge module and the solution module.
- Only the permitted axioms for that configuration are used:
-
["propext", "Quot.sound"](strictly constructive — noClassical.choice): enforced onconstructive.json,goldbach_21.json,zf_380.json, andriemann_104.json, covering all three ZF machine-to-proof-search equivalences (halts_zf393_iff_zfInconsistent,Zf380.Semantics.halts_candidate_iff_zfInconsistent,halts_zf432_iff_wadeZFInconsistent), first-orderPAconsistency and$\Sigma_1^0$ adequacy (not_paProvable_false,paHalts_holds_iff_halts,paZFInconsistent_holds_iff_zfInconsistent), the direct simulation reductions (Goldbach21,Zf380,Riemann104), theErdosTernary14reduction, all fivecandidate_fewer_statestheorems, and thebrocard_28equivalenceBrocard28.halts_brocard28_iff_exists_square(16 theorems total). DisallowingClassical.choiceblocks classicalCon(ZF)shortcuts (not_zfInconsistentinZFSet.{0}orClassical.em+digama0/ConZF) forHalts M ↔ ZFInconsistent, while deeply embeddedPAProvableprovides a weak first-order metatheory strictly belowZF(seeCryptids/Semantics/README.md). -
["propext", "Quot.sound", "Classical.choice"]: allowed inall.json,busy_beaver.json, and the classical model-theoretic / arithmetic-stage / analytic configs (pa_semantics.json,zf_393.json,zf_semantics.json,goldbach_semantics.json,riemann_95.json,riemann_120.json,riemann_semantics.json,erdos_ternary_15.json,erdos_ternary_14.json,brocard_28.json).
-
- The exported proof terms are accepted by the Lean default kernel.
| Config File | Challenge Module | Solution Module | Permitted Axioms | Verified Theorems |
|---|---|---|---|---|
constructive.json |
Challenge |
Cryptids |
propext, Quot.sound
|
All 16 constructive theorems: all 3 ZF proof-search equivalences (zf_393, zf_380, zf_432), first-order PA consistency & not_paProvable_false, paHalts_holds_iff_halts, paZFInconsistent_holds_iff_zfInconsistent), direct state reductions (Goldbach21, Zf380, Riemann104), Riemann95.candidate_fewer_states, and the ErdosTernary14 reduction, and Brocard28.halts_brocard28_iff_exists_square
|
all.json |
Challenge |
Cryptids |
propext, Quot.sound, Classical.choice
|
All 36 conjecture-checking state-reduction, equivalence, PA metatheory & internalized ZF provability, and semantics theorems |
pa_semantics.json |
Challenge |
Cryptids.Semantics.PA |
propext, Quot.sound, Classical.choice
|
not_paProvable_false, paHalts_holds_iff_halts, paZFInconsistent_holds_iff_zfInconsistent, paProvable_sound_nat, paProvable_halts_zf380_iff_zf393, paProvable_halts_zf393_iff_zfInconsistent, paProvable_halts_zf380_iff_zfInconsistent
|
busy_beaver.json |
BusyBeaverChallenge |
BusyBeaver |
propext, Quot.sound, Classical.choice
|
All 735 BB(6) and BB(3, 3) holdout Level A, Level B, equivalence, A6 is_not_halt, and hardness-evidence theorems (is_closed_pattern_set, has_collatz_reduction, is_closed_stack_pattern_set, has_stacked_collatz_reduction, has_stack_reduction, has_odometer_reduction, has_transducer_reduction, is_closed_segmented_system, is_valid_acceleration, isHalt_iff_acceleration, is_valid_highLevel_acceleration, isHalt_iff_highLevel_acceleration, acceleration_iff_highLevel_acceleration, is_not_halt, isHalt_HHH1LC_iff_HHH0LA, HardnessEvidence.*, is_word_run_cryptid, is_segmented_cryptid, is_odometer_cryptid, is_stack_cryptid, is_cryptid) |
a6.json |
BusyBeaverChallenge |
BusyBeaver.BB6.M_1RB0LF_1RC0RA_1LD0RB_0LE0LC_0LA0LC_0LBHHH |
propext, Quot.sound, Classical.choice
|
A6 is_not_halt
|
goldbach_21.json |
Challenge |
Cryptids.Goldbach21 |
propext, Quot.sound
|
candidate_fewer_states, candidate_equivalent
|
goldbach_semantics.json |
Challenge |
Cryptids.Semantics.Goldbach |
propext, Quot.sound, Classical.choice
|
halts_reference_iff_not_goldbach, halts_candidate_iff_not_goldbach
|
zf_380.json |
Challenge |
Cryptids.Zf380 |
propext, Quot.sound
|
candidate_fewer_states, candidate_equivalent
|
zf_393.json |
Challenge |
Cryptids.Semantics.Zf393 |
propext, Quot.sound, Classical.choice
|
halts_zf393_iff_zfInconsistent, not_halts_zf393, halts_candidate_iff_zfInconsistent, not_halts_candidate
|
zf_semantics.json |
Challenge |
Cryptids.Semantics.Zf |
propext, Quot.sound, Classical.choice
|
zfProvable_sound_zfSet, not_zfInconsistent, halts_zf432_iff_wadeZFInconsistent
|
riemann_104.json |
Challenge |
Cryptids.Riemann104 |
propext, Quot.sound
|
candidate_fewer_states, candidate_equivalent
|
riemann_95.json |
Challenge |
Cryptids.Semantics.Riemann95 |
propext, Quot.sound, Classical.choice
|
candidate_fewer_states, candidate_equivalent, halts_candidate_iff_not_riemannHypothesis
|
riemann_120.json |
Challenge |
Cryptids.Semantics.Riemann120 |
propext, Quot.sound, Classical.choice
|
halts_riemann120_iff_not_riemannHypothesis, halts_candidate_iff_not_riemannHypothesis
|
riemann_semantics.json |
Challenge |
Cryptids.Semantics.Riemann |
propext, Quot.sound, Classical.choice
|
halts_riemann744_iff_exists_counterexample, halts_riemann744_iff_not_riemannHypothesis
|
erdos_ternary_15.json |
Challenge |
Cryptids.Semantics.ErdosTernary15 |
propext, Quot.sound, Classical.choice
|
halts_erdosTernary15_iff_exists_counterexample, halts_erdosTernary15_iff_not_erdosTernary
|
erdos_ternary_14.json |
Challenge |
Cryptids.Semantics.ErdosTernary14 |
propext, Quot.sound, Classical.choice
|
candidate_fewer_states, candidate_equivalent, halts_candidate_iff_not_erdosTernary
|
brocard_28.json |
Challenge |
Cryptids.Semantics.Brocard28 |
propext, Quot.sound, Classical.choice
|
halts_brocard28_iff_exists_square, halts_brocard28_iff_not_brocard
|
The focused a6.json configuration checks the A6 IsNotHalt
theorem against BusyBeaverChallenge, using only its machine module as the
solution import. It uses the same three-axiom policy as the full Busy Beaver
configuration.
From the repository root (with bwrap installed):
lake exe cache get
lake comparator --config comparator/constructive.json
lake comparator --config comparator/all.json
lake comparator --config comparator/busy_beaver.json