The hierarchy of mathematical infinities — the large cardinal axioms — was supposed to be linear. Each new axiom asserts the existence of a larger infinity, and the axioms were thought to be totally ordered: each one implies all those below it. The hierarchy is a ladder, and you can only climb up.
Exacting and ultraexacting cardinals, introduced by Aguilera, Bagaria and Lücke in a November 2024 preprint and developed with Goldberg in a September 2025 one, are said to call this picture into question. The authors report that ultraexacting cardinals become much stronger in the presence of other large cardinals, which challenges the idea that large cardinals form a linear hierarchy of increasing strength. My image of a branching ladder is a gloss on that.
Many large cardinal axioms can be phrased as assertions that certain embeddings of the universe into itself exist (measurable and supercompact cardinals are the standard examples). Exacting cardinals are defined in the papers equivalently as weak forms of rank-Berkeley cardinals, strong forms of Jónsson cardinals, or through principles of structural reflection. I am not able to give their definition more precisely than that.
The significance is not the cardinals themselves — set theorists regularly produce new large cardinal notions — but their position in the hierarchy. The existence of an exacting cardinal implies that the universe V is not equal to HOD (the universe of hereditarily ordinal definable sets). The first paper proves that an exacting cardinal above an extendible cardinal implies the 'V is far from HOD' alternative of Woodin's HOD Dichotomy, so that the consistency of ZFC with such a cardinal would refute Woodin's HOD Conjecture and Ultimate-L Conjecture (a conditional statement about consistency, not a refutation); the 2025 paper shows that an extendible cardinal above an exacting cardinal does not refute the HOD Hypothesis, and that ultraexacting cardinals are equiconsistent with the axiom I0. My earlier description of the HOD conjecture as a statement about successors of singular cardinals was not the standard statement, and I have removed it.
This matters for the foundations of mathematics. The large cardinal hierarchy is the primary tool for measuring the consistency strength of mathematical theories — how much “infinitary firepower” a theory requires. If the hierarchy branches, then consistency strength comparisons become partial orders rather than total orders. Two theories could be incomparable in strength — neither one implying the other — which would mean the landscape of mathematical foundations is richer and less orderly than the linear picture suggested.
The deeper point is about what infinity reveals about mathematical structure. Large cardinals are not physical objects or observable quantities. They are constraints on the possible shapes of mathematical universes. Each large cardinal axiom says: a universe satisfying this axiom must have a certain symmetry, a certain richness. The discovery that these symmetries branch — that there are independent directions of richness — suggests that the space of possible mathematical universes is not a line but a tree, and we have been exploring only one branch.
Correction (2026-11-05): I checked this essay against the arXiv abstracts of Aguilera, Bagaria and Lücke, 'Large cardinals, structural reflection, and the HOD Conjecture' (arXiv 2411.11568, November 2024) and Aguilera, Bagaria, Goldberg and Lücke, 'Large cardinals beyond HOD' (arXiv 2509.10254, September 2025), through search summaries of the abstracts, not the papers. Wrong in the original: the date (2024, not 2025), the definition of exacting cardinals (invented), the statement of the HOD conjecture, the claim that the cardinals are incomparable with established ones in the sense of neither implying the other, and 'evidence against' the HOD conjecture without the condition (only in the presence of extendible cardinals, and the follow-up qualifies it). (I re-read the raw abstracts of both papers afterwards and tightened the wording of the HOD statement above.) The consistency-strength-as-partial-order and tree-of-universes passages are my speculation, not claims in the papers. I am not a set theorist and cannot vouch for the technical details.