Answers are grounded in this paper and cite their sources.
Status of this paper. This is a position paper. It contains no proof of P ≠ NP and does not claim to contain one. It contains a precise reformulation, one unconditional theorem, one proposition assembled from known results, a classification conjecture, a structural conjecture equivalent to P ≠ NP, and a ladder of fragments on which this conjecture can be attacked step by step. We reformulate the P versus NP question as a question about the existence of a uniform, polynomial-time section of propositional truncation, ‖A(φ)‖ ⊸ A(φ), for the family of 3-SAT witnesses, in a type system T that characterises the class P in the spirit of Soft Linear Logic (Lafont) and Light Linear Logic (Girard). We distinguish three levels at which a gap can appear between intension (the witness) and extension (the fact that a witness exists): logical, symmetric and resource. At the logical level there is no gap: for a finite, ordered domain the section exists in plain HoTT. At the symmetric level we prove unconditionally that no categorically natural section exists; we stress that this obstruction is purely symmetric, does not depend on resource bounds, and applies to 2-SAT as well. At the resource level we assemble known results into a proposition about the branching model L (self-reduction with unit propagation and a fixed order): the section L is polynomial for 2-SAT and Horn-SAT, but exponential for 3-SAT and, contrary to a naive reading of Schaefer's dichotomy, for XOR-SAT, under every fixed order. We propose that the boundary is drawn here not by polymorphisms but by propagation safety: we prove that it is sufficient and state its necessity as a conjecture. Ruling out a section in the whole of T is equivalent to P ≠ NP; we state it as a conjecture and propose a ladder of increasingly strong fragments as a programme for attacking it. We make the title formula precise through a parameter N, the number of witnesses. For N = 1 the logical and the symmetric gap vanish, while the Valiant–Vazirani theorem implies that the resource gap remains (unless NP = RP). Equality at two levels thus isolates the inequality at the third, and this is the exact content of the thesis “P ≠ NP precisely because P = NP”. Along the way we distinguish two inequalities: an informational one, the log₂ N bits erased by truncation, which vanishes at N = 1, and a computational one, which remains.
No takes yet. Share an insight, caveat, or question.
Jan Jagielski (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: