Episode 299, October 2026

No characters, then Z-stable?

An AI preprint claims the Kirchberg–Rørdam character criterion for Jiang–Su stability

Covers result 299 in OpenAI’s catalog: its one paper. See its connections to other results and the work it cites.

Play Episode 299: No characters, then Z-stable? Commutator bars halving along a chain of tensor legs.

3:42. English captions available. The video loads from YouTube when you press play.

Chapters

The claim, in plain words

A character of a C*-algebra is a one-dimensional representation, a homomorphism onto the complex numbers; matrix algebras of size two or more have none. Kirchberg’s central sequence algebra collects the sequences that asymptotically commute with the whole algebra, and 𝒵-stability, absorbing the Jiang–Su algebra, is one of the regularity properties from episode 291. In 2014 Kirchberg and Rørdam asked whether a unital separable algebra is 𝒵-stable exactly when its central sequence algebra has no characters; they called one direction, 𝒵-stable implies no characters, trivially true. They showed the question equivalent to one about infinite tensor powers, implicit in work of Dadarlat and Toms, and a problem list still carried it, unsolved, in May 2026.

Result 299 is one OpenAI paper, 21 pages, dated September 25, 2026, and published October 6. It claims that a nonzero unital separable C*-algebra is 𝒵-stable exactly when its central sequence algebra has no characters, for every free ultrafilter; the new part is the hard direction. Its key step: every nonzero unital algebra without characters receives a unital image of the dimension-drop algebra I(2,3) in some finite maximal tensor power of itself. So the infinite tensor power of such a separable algebra is 𝒵-stable, answering the Dadarlat–Toms form of the question.

What it doesn’t claim

That any finite tensor power is 𝒵-stable, a bound on the power, or anything for non-unital algebras.

Where it stands

Posted as a preprint: OpenAI, 21 pages, dated September 25, 2026
Peer reviewed: Not yet
Formalized in Lean: In part: the main criterion, not the tensor-power result (docs)

Last checked October 19, 2026. The paper has not been revised. In Lean, the main criterion is formalized and matches the paper, with its own model of the Jiang–Su algebra, in about 25,000 lines, with no unproved steps in our scan; an outside project re-ran the proof against the formal statement. The tensor-power result is not formalized. OSVerify has not reviewed it, and we found no expert comment.

How the proof goes

The target. The dimension-drop algebra I(2,3) has irreducible representations of sizes 2, 6 and 3, and no characters; Kirchberg and Rørdam had shown that a unital map of it into the central sequence algebra is all that is needed.
A cube. The paper builds the map on a cube of coordinates, sliding a copy of the 2×2 matrices from one tensor leg to the next. The cost is fast-growing exponents: for a comparison constant of 3 the bookkeeping uses 11 copies of an earlier tensor power, and for 10, 10,004.

The numbers in the video

Ours: −(i/2)[σ_x, σ_y]σ_z = 1, so 2×2 matrices have no character; I(2,3) has irreducible representations of sizes 2, 6 and 3 and no character, while I(1,3) has one.
Ours: in ten tensored copies of 2×2 matrices, σ_z on leg k has commutator 2^1−k with a fixed element whose weights halve along the legs.
Ours: the paper’s cube construction checked at sample points to about 10⁻¹⁵.

Read the source

The Kirchberg–Rørdam Character Criterion (PDF), 21 pages.
Kirchberg and Rørdam, “Central sequence C*-algebras and tensorial absorption of the Jiang–Su algebra” (arXiv, 2014).
The Lean formalization notes.
The collection’s catalog, where this is result 299.

The paper

OpenAI groups related papers into one result; this one has a single paper.

The Kirchberg–Rørdam character criterion

Transcript

Read the full script

0:00The question

A character of a C*-algebra is a one-dimensional representation, a homomorphism onto the complex numbers. Matrix algebras of size two or more have none: our computer checks that −i/2, times the commutator of two Pauli matrices, times the third, is the identity, so a character would have to send 1 to 0. A direct sum of matrix blocks has exactly one character for each 1×1 block. Kirchberg’s central sequence algebra collects the sequences that asymptotically commute with the whole algebra. Our computer shows them in ten tensored copies of 2×2 matrices: a copy of σ_z placed further along commutes better and better with a fixed element whose weights halve along the legs, while two such sequences need not commute with each other. Absorbing the Jiang–Su algebra, 𝒵-stability, is one of the regularity properties from episode 291. In 2014 Kirchberg and Rørdam asked whether a unital separable algebra is 𝒵-stable exactly when its central sequence algebra has no characters; they called one direction, 𝒵-stable implies no characters, trivially true. They showed the question equivalent to one about infinite tensor powers, implicit in work of Dadarlat and Toms, and a problem list still carried it, unsolved, in May 2026.

1:23The claim

Result 299 is one OpenAI paper, 21 pages, dated September 25 and published October 6, 2026. It claims that a nonzero unital separable C*-algebra is 𝒵-stable exactly when its central sequence algebra has no characters, for every free ultrafilter. The new part is the hard direction: no characters implies 𝒵-stable. Its key new step: every nonzero unital algebra without characters receives a unital image of the dimension-drop algebra I(2,3) in some finite maximal tensor power of itself. So the infinite tensor power of such a separable algebra is 𝒵-stable, answering the Dadarlat–Toms form of the question. Not claimed: that any finite tensor power is 𝒵-stable, a bound on the power, or anything for non-unital algebras.

2:17How the proof goes

The dimension-drop algebra I(2,3) has irreducible representations of sizes 2, 6 and 3, and no characters; Kirchberg and Rørdam had shown that a unital map of it into the central sequence algebra is all that is needed. Our computer checks those sizes, and that I(1,3), by contrast, has a character. The paper builds the map on a cube of coordinates, sliding a copy of the 2×2 matrices from one tensor leg to the next; our computer checks that construction at sample points, to about 10⁻¹⁵. The cost is fast-growing exponents: for a comparison constant of 3 the bookkeeping uses 11 copies of an earlier tensor power, and for 10, 10,004.

3:02Where it stands

The paper is unrefereed and unchanged since release. In Lean, the main criterion is formalized and matches the paper, with its own model of the Jiang–Su algebra, in about 25,000 lines; our scan finds no unproved steps, and an outside project re-ran the proof against the formal statement, but the tensor power result is not formalized. We found no expert comment. If it holds, a unital separable algebra absorbs the Jiang–Su algebra exactly when its central sequence algebra has no characters, with no further hypotheses.

Spot a mistake?

This episode simplifies on purpose, but it shouldn’t be wrong. If something is, open an issue on GitHub or leave a comment on YouTube. Corrections are noted here with a date.