Bonjour! I am Erik and this is my home on the interwebs. Here you will find a random collection of things that I have been working on.
The rewrite system of a circuit can prove two circuits equal — but only exactly, and most useful transformations of a physical system are approximate. Simulation learning weakens equivalence the way Hoare logic weakens pre- and postconditions: an agent interleaves exact symbolic rewrites with approximate ones justified by simulation, producing a mixed proof that carries an explicit error budget.
Read more
An agent finishes a task and forgets almost everything. Dreaming is an offline pass that replays the scattered notes an agent saved, distils them into durable lessons, and folds those lessons back into its prompt — so it wakes up knowing what it learned yesterday. Self-improvement with no retraining.
Read more
Why does next-token prediction learn language so well, yet stall on dynamical systems? An explanation rooted in the smoothness of the map from syntax to meaning — and what it implies about where each domain earns its competence: pre-training, reinforcement learning, or search.
Read more
A programming language for dynamical systems — a LEAN-style parser, a rewrite engine, a circuit compiler, and a JAX runtime. The mathematical structure stays explicit from equation to execution, enabling formal composition, algebraic rewriting, and interchangeable semantics.
Read more
Most agent frameworks treat LLMs as agents. Hugin treats them as oracles — one component in a larger reasoning system. Built around an immutable stack architecture that makes branching, debugging, and multi-agent coordination natural.
Read more
This is part four in a series on how to use a state machine framework to model agentic flows and how this approach enables some interesting features. In this part we explore advanced techniques for improving multi-agent reasoning.
Read more
In the last two parts we described the state machine framework, how it works and how we can steer agents within it. In this third part, we will describe several other desirable extensions of any agent framework and how we can implement them with our state machine setup.
Read more
In the first part we described the state machine framework and how it works. In this part we will dig into some of the techniques for steering agents and how we implement them in the state machine framework.
Read more
Since I last wrote a few thoughts down on agents the buzz has only increased. With our release of Bigwig I thought it would also be interesting to dig into some of the details of how we have gone about implementing our system of multi-agents.
Read more
Recently Yoav Goldberg asked a series of interesting question on Bluesky(1, 2, 3).
Read more
I think this is one of the more interesting and peculiar operators in matrix algebra, so I wanted to write a little bit about it.
Read more
I have recently been thinking about modelling of dynamical systems again - so wanted to revisit some of my old ideas.
Read more
At the end of October, I gave a quick demo at AI Tinkerers Paris about one of our early, early version of BigWig. This is a short writeup based on the slides, called "Agents for building ML models".
Read more