This essay offers a way to see three non-classical logics — type theory /intuitionism, quantum logic, and Zero-Trust Logic (ZTL) — as threedisciplined answers to a single self-referential paradox. The paradox is theone construction behind the liar, Russell's set, Cantor's diagonal, andGödel's undecidable sentence, made exact by Lawvere's fixed-point theorem(1969): self-application with no fixed point. Each keeps all of classical logicexcept one law and thereby houses the paradox differently FROM WITHIN — typetheory by the universe hierarchy (Girard's paradox), quantum bysuperposition, ZTL by pointwise quarantine to the mark Z. Classical logicitself is stated once as the BASELINE and set aside: it keeps everystructural principle and therefore has nothing to house the paradox with, soit explodes and is repaired from the METALANGUAGE — Tarski's hierarchy sitsabove the logic, not inside it. Its "resolution" is of a different kind, andnothing about the classical corner is proved here. Two invariants do not move: non-contradiction with its reductio(the shared floor, a law kept by all) and the self-referential paradox itself(the shared ceiling, a monster tamed by all); the floor is the guard postedagainst the monster. The three do not merge, and the paper states exactly how the three pairsfail to: two pairs clash at the level of what is validated (total excludedmiddle against its pointwise failure; strong normalization against housednon-termination), while the third — quantum and intuitionistic — CAN merge,and the merge is exactly classical logic (distributivity plus excluded middle,double negation and total complementation is a Boolean algebra). Mergingexpels the pair from the family: the result resolves nothing from within. Sothe three are incomparable AS RESOLUTIONS — two pairs cannot share a valuationstructure, and the third can share one only by ceasing to resolve. The formal components are machine-checked on the EMPTY axiom list,verifiable from zero with a bare Lean 4 (no mathlib, no imports): themirror between ZTL and quantum logic, the combinatorialcore of quantum contextuality — the Mermin-Peres magic square admits nobivalent valuation (0 of 512) and GHZ none (0 of 64), by kernelenumeration. On two quanta the mirror sharpens: at the singlet, pairpropositions are true while every local proposition is empty — onticvacancy, not our ignorance: by Bell no consistent local values exist to belooked at — the covering law falls in correlation form — and the seam itself iskernel-checked (the junction theorem: the singlet lies in the join of twoproduct atoms yet in neither atom and in no local plane of either factor,exact integer arithmetic, empty axiom list); the falls saturate along theladder. A closing section, explicitly marked as a reading, draws theconsequence: the corners are descriptions and it is descriptions that paythe laws; at the level of states the diagonal's premise itself fails (theself-negating fixed point exists physically), so for the world the paradoxdissolves rather than being tamed. ZTL keeps distributivity and losesexcluded middle, double negation, and identity p→p (they fall at the mark);quantum logic (witnessed on MO2, the smallest non-distributive ortholattice)keeps excluded middle and double negation and loses distributivity; both keepnon-contradiction. ZTL breaks where a thing equals itself; quantum breaks wherethings combine. That second clause is sharpened into a necessity andmachine-checked: MO2 keeps modus ponens under the Sasaki hook but admits NOimplication at all — no binary operation on the lattice — satisfying thededuction theorem: such an arrow would have to be a relativepseudocomplement, which would make the lattice Heyting andtherefore distributive. So what the quantum corner cannot do is discharge apremise while a context still stands, and no choice of connective repairs it;the claim is not "MO2 under the Sasaki hook lacks the deduction theorem" but"MO2 cannot have one". The mirror is shown to be an analogy, not a lattice duality:the obstruction is the involution asymmetry (ZTL's negation is not involutive,double negation being one of the laws it drops), which is also what locatesZTL as the paracomplete relative that breaks the involution its Lukasiewiczancestor and its quantum cousin both keep (the known bridge from orthomodularto many-valued logic runs through Lukasiewicz, Pykacz 2010). What changed in version 1.1.0. Classical logic is no longer listed as afourth resolution. Version 1.0.0 tabled it beside the other three, whichcounted classical logic among the "non-classical logics" of its own subtitle,and suggested a symmetry that does not hold: none of the five machine-checkedcomponents concerns the classical corner — every witness is about ZTL, MO2, ortheir pair. Classical logic is now stated once as the baseline and set aside,with the difference in kind made explicit (its repair is metalinguistic, theother three resolve the diagonal from within). Section 4 is restated exactly: two pairs are valuation-incompatible, and thethird pair (quantum–intuitionistic) can merge — into classical logic itself,which is precisely the baseline and resolves nothing from within. The oldargument through the baseline was muddled in form (classical logic is nottrivial) but was gesturing at this true fact, and the repair states it ratherthan deleting it: the baseline is where the mergeable pair lands. The metaphorof an orbit around an unreachable centre is gone; it was marked as metaphorand did no work. No formal content is added or removed: the fivemachine-checked components, their axiom profiles and their statements areunchanged from 1.0.0. Honest scope. This is a synthesis and a reading, not a theorem, a merger, a newfoundation, or a new field. Its components are established (Lawvere 1969;Birkhoff–von Neumann 1936; intuitionism; universal logic). What is the author's:ZTL itself (its own preprint), the curation of the three as resolutions of onediagonal, and the five Lean witnesses (the impossibility result among them). The formal components are machine-checked;the cycle between them, and the metatheoretic corners (intuitionisticunderivability of LEM), are prose, marked as such. Thereliability of the machine-checked components does not depend on trusting theauthor or the AI: all four Lean files verify against the Lean 4 kernel on theempty axiom list, 35 objects in about a second in total. AI disclosure: written in dialogue with the AI system Claude (Anthropic); alldesign decisions, framing, and final responsibility rest with the human author.
Vitaliy Reznik (Mon,) studied this question.