Uploaded September 2026 | Updated November 2026, 2 hours ago
Title: Organising Compositionality in Classical Mechanics
Abstract: Consider a scenario where we give two people each two springs and ask them to build something with them. There is no reason to believe that the resulting systems would be the same — one may choose to make the springs share a common mass, while another may decide to leave them separate. While the components remain the same, the resulting system does not. Mathematically, where exactly should the choice of how to put things together live?
Baez, Weisbart and Yassine think about this by modelling open classical systems as augmented spans and composing compatible systems using pullbacks. Their construction keeps track of Lagrangian and Hamiltonian descriptions of mechanics — the former using positions and velocities, the latter positions and momenta — but doing so requires a fair amount of machinery, including separate Riemannian and Poisson settings and a generalized F-pullback construction. I’ll reorganise this picture by working with smooth spaces carrying closed 2-forms, where ordinary pullbacks are available. These 2-forms encode the geometry of position–momentum phase space, and our systems are required to respect that geometry. For the spring examples, differentiating the potential picks out a half-dimensional subspace on which the 2-form vanishes — called a Lagrangian submanifold. The resulting span theory then keeps the systems themselves separate from the ways of assembling them. So we can fix two springs once and vary only the instruction for how they should be combined.
Most of the talk will involve seeing this in action through examples involving springs. Tensoring two spring systems simply places them side by side, without yet deciding how they are connected. That extra choice lives in a loose span, whose action by pullback performs the assembly. Changing the span while keeping the springs fixed can therefore produce different mechanical systems. We’ll work this out explicitly for two springs sharing a mass, and end by asking whether the same mechanism can also compose Hamiltonian evolution through time.
Date: 8/25/26
Title: Organising Compositionality in Classical Mechanics
Abstract: Consider a scenario where we give two people each two springs and ask them to build something with them. There is no reason to believe that the resulting systems would be the same — one may choose to make the springs share a common mass, while another may decide to leave them separate. While the components remain the same, the resulting system does not. Mathematically, where exactly should the choice of how to put things together live?
Baez, Weisbart and Yassine think about this by modelling open classical systems as augmented spans and composing compatible systems using pullbacks. Their construction keeps track of Lagrangian and Hamiltonian descriptions of mechanics — the former using positions and velocities, the latter positions and momenta — but doing so requires a fair amount of machinery, including separate Riemannian and Poisson settings and a generalized F-pullback construction. I’ll reorganise this picture by working with smooth spaces carrying closed 2-forms, where ordinary pullbacks are available. These 2-forms encode the geometry of position–momentum phase space, and our systems are required to respect that geometry. For the spring examples, differentiating the potential picks out a half-dimensional subspace on which the 2-form vanishes — called a Lagrangian submanifold. The resulting span theory then keeps the systems themselves separate from the ways of assembling them. So we can fix two springs once and vary only the instruction for how they should be combined.
Most of the talk will involve seeing this in action through examples involving springs. Tensoring two spring systems simply places them side by side, without yet deciding how they are connected. That extra choice lives in a loose span, whose action by pullback performs the assembly. Changing the span while keeping the springs fixed can therefore produce different mechanical systems. We’ll work this out explicitly for two springs sharing a mass, and end by asking whether the same mechanism can also compose Hamiltonian evolution through time.
Date: 8/25/26


![[DOTS Lectures] 4. Composing Moore Machines
Part of a lecture series on the Double Operadic Theory of Systems (DOTS) presented by David Jaz Myers. [DOTS Lectures] 4. Composing Moore Machines](https://i.ytimg.com/vi/jsM2jTuT_Zc/mqdefault.jpg)
![[2-torial] David Jaz tells Brendan about a topos-theoretic interpretation for conceptual modelling
Recorded at the Oxford Office on the 5th of December 2025. [2-torial] David Jaz tells Brendan about a topos-theoretic interpretation for conceptual modelling](https://i.ytimg.com/vi/kFQpKp-ehZI/mqdefault.jpg)
![[Berkeley Seminar] Evan Patterson (Topos Institute) | A uniform type theory for internal languages
Title: A uniform type theory for internal languages of substructural categorical logics
Abstract: Bridging between category theory and type theory, the internal language of a categorical structure enables morphisms to be specified in an ergonomic notation based on elements and function application. Finding an internal language is typically a manual process, performed anew for each categorical structure. We propose a uniform type theory giving internal languages of many different categorical structures corresponding to substructural logics, some familiar and others less so. Structures covered include categories, multicategories (planar, symmetric, and cartesian), PROPs, Lawvere theories, (symmetric) monoidal categories, cartesian categories, Markov categories, and categorical notions of database schema. We present the main ideas and selected examples of this type theory, and we preview an implementation of the type theory by our colleague, tslil clingman.
Date: Sept 29, 2026
Link to slides: https://www.epatters.org/assets/slides/2026-09-berkeley-seminar/slides.html [Berkeley Seminar] Evan Patterson (Topos Institute) | A uniform type theory for internal languages](https://i.ytimg.com/vi/kPLiC8zUKLA/mqdefault.jpg)

![[DOTS Lectures] 1. Categories of systems
Part of a lecture series on the Double Operadic Theory of Systems (DOTS) presented by David Jaz Myers. [DOTS Lectures] 1. Categories of systems](https://i.ytimg.com/vi/kZ4muU5Wc_4/mqdefault.jpg)
![[Berkeley Seminar] Raph Levien | How Rust won: the quest for performant, reliable software
Title: How Rust won: the quest for performant, reliable software
Abstract: For a long time, high performance has been in tension with reliability. In particular, languages designed for high performance were not memory safe, with real implications for unexpected crashes and security vulnerabilities. Rust is the first practical language to address this tension, building on affine types and other principles programming language theory, synthesized with an attention to low level systems programming. Rust’s success emerged not just from a clever idea, but consistently excellent execution and the formation of a strong community around the language. This talk will discuss several aspects of what Rust got right, as well as the rocky journey of ideas from academic theory to real world impact.
Slides: https://docs.google.com/presentation/d/1SoDsm_m_pb_gS6Y98HghhzBviYZxp3F2XawhIppJQQo/edit?usp=sharing
https://topos.institute/events/berkeley-seminar/ [Berkeley Seminar] Raph Levien | How Rust won: the quest for performant, reliable software](https://i.ytimg.com/vi/k_-6KI3m31M/mqdefault.jpg)
![[DOTS Lectures] 11. A general representability theorem for Systems Theory Pt. 1
Part of a lecture series on the Double Operadic Theory of Systems (DOTS) presented by David Jaz Myers.
Some material from these lectures can be found in Davids book on categorical systems theory:
https://www.davidjaz.com/Papers/DynamicalBook.pdf [DOTS Lectures] 11. A general representability theorem for Systems Theory Pt. 1](https://i.ytimg.com/vi/kpmhI1oHGnM/mqdefault.jpg)

![[TopOx] Steve Awodey: Path Types in Algebraic Type Theory
18th of May 2026. Slides available at https://topos.institute/events/topox/
A representable natural transformation u : U* → U in the category Psh(C) of presheaves on a small category C is a “natural model of dependent type theory. The type-forming operations may be described as an algebraic structure on u, representing corresponding operations on the type-families classified by u. For example, the dependent product or “Pi-type” is an algebra structure for the polynomial endofunctor
P_u : Psh(C) → Psh(C) .
Similar operations on u represent the other type-formers of unit type, dependent sums, and identity types. The latter are given by a recently determined “path-type” structure, which relates such models to cubical (Quillen) model categories. [TopOx] Steve Awodey: Path Types in Algebraic Type Theory](https://i.ytimg.com/vi/lanvZuki4qQ/mqdefault.jpg)