Recorded description
CReal.lub_decides_em is the least-upper-bound property's ADR-0603 row 2 (Spivak Calculus ch. 8), the row docs/curriculum/graded-statement-families.md recorded on 2026-08-27 as PURE ABSENCE -- 'the unavailability is asserted, not proved'. Spivak's P13 quantifies over an arbitrary inhabited set of reals that is bounded above, so the faithful counterexample family is a set carved out by an arbitrary proposition: CReal.lubSet A := fun x => Or (le x zero) (And A (le x one)), the set (-inf, 0] union ((-inf, 1] if A). It is inhabited at 0 (CReal.lubSet_inhabited, an exhibited witness rather than an Exists) and bounded above by 1 (CReal.lubSet_bounded, an explicit bound rather than an Exists), both PROVED here rather than asserted, so the family is machine-checked to lie inside classical LUB's hypothesis class. Given a supremum s in BISHOP's sense -- an upper bound plus the approximation property, for every t < s some x in the set exceeds t -- one CReal.lt_cotrans call on the fixed strict pair zero < one at z := s yields Or A (Not A) for an ARBITRARY Prop. That is UNRESTRICTED EXCLUDED MIDDLE, which this kernel does not contain: only Decidable.em, which takes a Decidable instance, and the conditional em_of_dne / em_of_peirce / dne_of_em / peirce_of_em, which take unrestricted em as a HYPOTHESIS and never assert it (ADR-0716 section 2 measures that absence with controls). So this is a STRICTLY STRONGER boundary than F:creal-evt-attained-max-decides-sign and F:creal-ivt-exact-root-decides-sign, which both land on analytic LLPO -- a principle consistent with Bishop's constructive mathematics, where unrestricted excluded middle is not. What this is NOT: a proof that Or A (Not A) is false. Excluded middle is consistent with this kernel's type theory, so the conclusion is UNPROVABLE here, not refutable; 'refuted' means exactly what it means for the IVT and EVT rows -- the classical conclusion is proved at least as strong as a decision principle this kernel demonstrably does not have. It is falsifiable: land an unrestricted em and this stops being a refutation and becomes a route to LUB. Nor does it contradict LUB's row 1: CReal.supOn / CReal.supOn_approx_lub (F:creal-supon, F:creal-supon-approx-lub) construct the supremum of a UNIFORMLY CONTINUOUS FUNCTION on a compact interval, where the modulus supplies the locatedness a general set lacks, and creal/completeness.rs's Bishop completeness constructs the limit of a REGULAR SEQUENCE, which carries its own rate. lubSet A is a set, not a function's range and not a regular sequence, and it is exactly as un-located as A is undecided.