DontFearTheProfunctorOptics VS cofree-bot

The best one I know of is this: DontFearTheProfunctorOptics. Hopefully someone else has it translated to other languages - I guess typeclass would be a huge bottleneck.

Profunctor optics and the journey to learn the differences between the representations will clear things up. https://github.com/hablapps/DontFearTheProfunctorOptics/blob/master/Optics.md is a great read. The Profunctor have more flexible composition which is both a good and bad thing and is still debated. But understanding them and the notions of Profunctors generally applied to a domain will align your current natural intuition.

Not a job per se, but some friends and I wrote a library for building compositional chatbots (like for IRC not LLMs necessarily) encoded as Mealy Machines: https://github.com/cofree-coffee/cofree-bot/

That project then has since led to a long term collaboration with the Topos Institute where we are building type theory for Polynomial Functors: https://github.com/toposInstitute/polytt

Polynomial Functors are a really powerful abstraction from Category Theory which subsumes the co-algebraic approach to finite state machines used in `cofree-bot` and which can also be used to encode wiring diagrams, tactics engines, game semantics, neural networks, and dynamical systems in general.