Observations
on bricks in representation theory of algebras
This page collects observations on bricks — examples, counterexamples and short results that have been communicated to the Brick Lodge — with full credit to those who found them. The entries are listed in the order in which they were added, and may not yet have appeared in a refereed journal or on the arXiv.
Some of the observations below may be AI-assisted, or even entirely AI-generated, under a mathematician’s direction. They are shared solely to help build intuition and insight into the open problems of the Brick Lodge and related topics. If you wish to cite or use any of them in a publication or presentation, please contact the contributors directly. Each observation is credited to the person or people who communicated it.
The Brick Lodge welcomes feedback and comments from all experts on these observations, as well as suggestions for improving their exposition (with credit to the original work). To leave a comment, use the “Comments & Feedback” button at the bottom right of the relevant observation. Once the maintainer or one of the contributors of the Brick Lodge has reviewed and approved it, your comment will appear beneath that observation under your full name.
1. On infinite semibrick of pairwise extension-orthogonal rigid bricks
Setting: k is an algebraically closed field of characteristic zero, and A is a finite-dimensional k-algebra. A semibrick is a set of pairwise Hom-orthogonal bricks, and a brick X is rigid if Ext1A(X, X) = 0. A semibrick 𝕊 consists of pairwise extension-orthogonal rigid bricks if Ext1A(X, Y) = 0 for all X, Y ∈ 𝕊, including the case X = Y.
A finite-dimensional algebra can admit an infinite semibrick of pairwise extension-orthogonal rigid bricks.
Credit, background and motivation. This example was communicated to us by Haruhisa Enomoto, to whom we are grateful, in October 2026 (the attached note is dated October 4, 2026). Semibricks, which are in bijection with wide subcategories, are at the heart of the brick-Brauer–Thrall conjectures: the Semibrick Conjecture of Enomoto (arXiv:2005.01626) predicts that an algebra is brick-infinite if and only if it admits an infinite semibrick. It is natural to ask how well-behaved such a semibrick can be made. In arXiv:2606.27063 Mousavand–Paquette asked (Question 4.2) whether, over a fixed algebra, a semibrick of pairwise extension-orthogonal rigid bricks is necessarily finite and, if so, whether its cardinality is bounded in terms of the algebra. Their Example 4.1 produces finite such semibricks of arbitrarily large cardinality, but only by changing the algebra each time, and their argument (based on τ-tilting theory) shows that an algebra with an infinite such semibrick must have global dimension at least two.
Main result. Over an algebraically closed field of characteristic zero, there is a connected basic algebra Λ with three simple modules, of dimension 21, with (rad Λ)3 = 0 and global dimension 2 (the least possible), together with modules Bn (n ≥ 0) of dimension vector (n+1, n+1, n+1) such that HomΛ(Bm, Bn) is k if m = n and 0 otherwise, and Ext1Λ(Bm, Bn) = 0 for all m, n ≥ 0. Hence {Bn} is an infinite semibrick of pairwise extension-orthogonal rigid bricks, and over this one algebra even the sizes of finite such semibricks are unbounded. Each Bn has projective dimension exactly 2, and dim Ext2(Bm, Bn) = (m+1)(n+1) − δmn.
Construction and AI provenance. The algebra Λ is an instance of the universal-localization construction of Neeman–Ranicki–Schofield (Math. Proc. Cambridge Philos. Soc. 2004), applied to the universal enveloping algebra U of the Lie algebra sl2. The finite-dimensional U-modules embed fully and exactly into mod Λ with an extension-closed image (the modules on which two distinguished arrows act invertibly), so Hom and Ext1 are preserved, and Bn is the image of the (n+1)-dimensional simple sl2-module; the required vanishing then reduces to the complete reducibility of sl2-modules in characteristic zero. The example was created by AI: according to the note, the construction and its proof were developed with OpenAI’s Codex agent in research directed by Haruhisa Enomoto, and the manuscript — a LaTeX/PDF note titled “An infinite semibrick of rigid modules”, listing “Codex (GPT-6)” as its author — was drafted with GPT-6 in Codex. It has not yet been refereed.
Connections to other entries of the Brick Lodge:
• Open Questions, item 3 (Rigid semibricks): part (1) is answered affirmatively by this example; part (2) — whether such an algebra must be (strictly) wild — remains open. Our own remark: the quotient of Λ by the ideal generated by the idempotent of the third vertex is the path algebra of the Kronecker quiver with four arrows, which is strictly wild, so Λ is itself strictly wild; the example is therefore consistent with an affirmative answer to part (2). It also settles the existence version of the MathOverflow question.
• Semibrick Conjecture and 2nd bBT Conjecture: Λ is brick-infinite and admits an infinite semibrick, as predicted. The members Bn have pairwise different dimension vectors, but, being strictly wild, Λ also has infinitely many bricks of one dimension vector (arXiv:2606.27063, Proposition 3.1) — concretely, the bricks of dimension vector (1, 1, 0) given by the nonzero scalar 4-tuples on the four arrows from vertex 1 to vertex 2. Likewise, the family does not by itself witness condition (1) of the Hom-orthogonal Conjecture, which asks for pairwise Hom-orthogonal modules of the same dimension. However, one can also generate such a set.
• Rigid-bricks Conjecture: not affected. That conjecture assumes that every brick is rigid, whereas here only the members of the semibrick are; indeed the bricks of dimension vector (1, 1, 0) above are not rigid (already over the Kronecker quotient their self-extension space is 3-dimensional, and it embeds into that over Λ).
• The example is strictly wild and not brick-tame, so it does not bear on Questions 1 and 2 (tameness; no-gap phenomenon). We do not know of a direct implication for, or further insights into, the remaining questions and conjectures in the Brick Lodge.
Related topics
2. On Homological conjectures and boundedness conjecture
Setting: A is an Artin algebra, and modules are finitely generated. A module M is self-orthogonal if ExtiA(M, M) = 0 for all i ≥ 1. The Auslander–Reiten conjecture asserts that if ExtiA(M, M ⊕ A) = 0 for all i ≥ 1, then M is projective.
There exists an explicit algebra A and a non-projective module M with ExtiA(M, M ⊕ A) = 0 for all i ≥ 1; that is, a counterexample to the Auslander–Reiten conjecture.
The comment form has not been activated yet. Please check back soon, or contact Kaveh Mousavand directly.
Credit, background and motivation. This example was communicated to us in October 2026 by Haruhisa Enomoto and René Marczinzik, to whom we are grateful. It was constructed by GPT-6-Astra (OpenAI Codex, GPT-6) under their supervision, and the interactive page presenting it was made with Claude. The Auslander–Reiten conjecture, which dates back to the 1970s, is one of the central open problems of the homological theory of finite-dimensional algebras. Enomoto and Marczinzik have worked extensively on the surrounding circle of homological conjectures, including the recent preprint discussed below, so this is a natural place for such an example to appear. At the time of writing there is no arXiv preprint and the example has not been refereed.
Main result. There exists an explicit algebra, whose quiver, relations and module are given on the interactive page, that is a counterexample to the Auslander–Reiten conjecture: it carries a non-projective module M with Exti(M, M ⊕ A) = 0 for all i ≥ 1.
The boundedness conjecture. In Section 5 of “Maximal self-orthogonal modules and a new generalization of tilting modules” (Enomoto, arXiv:2301.13498) the boundedness conjecture appears as Conjecture 5.1. Roughly, it predicts that over a fixed algebra the number of pairwise non-isomorphic indecomposable summands of a self-orthogonal module is bounded by the rank of the algebra, that is, by the number of its simple modules. Remark 5.6 of that paper says that the conjecture appears in various publications and that its name originates from Happel: D. Happel, “Selforthogonal modules”, in Abelian groups and modules (Padova, 1994), Math. Appl. 343, Kluwer, 1995, pp. 257–276; and D. Happel and L. Unger, “Complements and the generalized Nakayama conjecture”, in Algebras and modules II (Geiranger, 1996), CMS Conf. Proc. 24, Amer. Math. Soc., 1998, pp. 293–310.
Boundedness and the homological conjectures. The boundedness conjecture implies the Auslander–Reiten conjecture and, globally, the generalized Nakayama conjecture. The first implication is short. Suppose M is non-projective with Exti(M, M ⊕ A) = 0 for all i ≥ 1. Then M ⊕ A is self-orthogonal, and since M has an indecomposable non-projective summand it has strictly more than n pairwise non-isomorphic indecomposable summands, where n is the number of simple modules, contradicting boundedness. Consequently, a counterexample to the Auslander–Reiten conjecture, such as the one above, is also a counterexample to boundedness: adjoining the regular module to a non-projective module that violates the Auslander–Reiten conjecture already gives a self-orthogonal module with too many indecomposable summands. Boundedness could fail much more strongly, for instance if one algebra had self-orthogonal modules with arbitrarily many summands. Observation 1 points in that direction but does not reach it: its infinite semibrick is only extension-orthogonal in degree one (Ext2 does not vanish there), so it is not self-orthogonal in the above sense. A semibrick of bricks with all higher extensions vanishing would give arbitrarily large ranks over one fixed algebra.
Connection to recent work on homological conjectures. In “On the homological conjectures for Artin algebras” (Enomoto–Marczinzik, arXiv:2609.19172, September 2026) the authors show that, for Artin algebras over a fixed commutative artinian ring, the Auslander–Reiten conjecture, the generalized Nakayama conjecture, the Nakayama conjecture and both Tachikawa conjectures are globally equivalent; in other words, they all hold or all fail. Taken together with the example above, and provided it is confirmed, this suggests that these conjectures fail as a family rather than just one of them. We stress that the equivalence is global: a counterexample to one conjecture yields counterexamples to the others over the same base ring, but not necessarily for the same algebra. We also stress that the equivalence is a statement about these conjectures; it does not itself say whether the boundedness conjecture is among them.
Connection to recent AI results on the finitistic dimension. OpenAI has recently released a collection of mathematical manuscripts (github.com/openai/math), among them a manuscript, “An algebra of infinite little finitistic dimension” (September 23, 2026), which, according to its accompanying Lean-formalization page, gives a finite-dimensional complex algebra with finitely generated modules of finite but unbounded projective dimension, and is presented as a disproof of the little finitistic dimension conjecture. The two results are logically independent, as far as we can tell: the finitistic dimension conjecture is about bounding projective dimensions, while the Auslander–Reiten and boundedness conjectures concern Ext-orthogonality and the number of summands. They are nonetheless conceptually close. All of them belong to the same family of “finiteness and boundedness” statements for finite-dimensional algebras that were long open, and both counterexamples were found with the help of AI systems. We do not know of a direct implication in either direction, and we would welcome comments from experts on this point.
Connections to other entries of the Brick Lodge:
• Observation 1: also produced with AI, and also concerned with Ext-orthogonality in a fixed algebra; see the remark above on why it falls short of self-orthogonality.
• We do not know of a direct implication for, or further insights into, the questions and conjectures of the Brick Lodge, which concern bricks, semibricks and brick-finiteness. Comments on possible connections are welcome.
Related topics
The comment form has not been activated yet. Please check back soon, or contact Kaveh Mousavand directly.