World-Models on Step Into Dev

The World Model Equation Is a WHERE Clause

Tue, 07 Jul 2026 00:00:00 +0000

The first code post in this series opened with a jab at explanations that lead with “a conditional probability, a subscript, a latent variable, and half the audience gone by paragraph three”. I also promised the maths would eventually arrive, once it described something you’d already watched run. This post makes good on that: it takes exactly those three scary things, one at a time, and shows that you’ve been writing all of them in TypeScript for years under different names.

A World Model in Forty Lines of TypeScript

Mon, 06 Jul 2026 00:00:00 +0000

Every explanation of world models I’ve read starts the same way. A conditional probability, a subscript, a latent variable, and half the audience gone by paragraph three. Which is a shame, because the core idea fits in a few lines of code and most working programmers already have the intuition. They’ve just never had it pointed at.

So this series starts with TypeScript. We’ll get to the equations eventually, but only once they’re describing something you’ve already watched run. (If you’d rather have today’s ideas with no code at all, the previous post covers the same ground in prose.)

Your Time Series Is Already Training Data

Sun, 05 Jul 2026 00:00:00 +0000

At Unimo we spend our days watching telemetry from energy hardware. Temperatures, states of charge, power flows, a reading every few seconds from devices sitting in buildings we’ll never visit. Alongside the readings sits a second stream that gets far less attention: the commands. Turn this on. Back that off. Hold here.

For years I thought of those two streams as operations data. You keep them so you can draw dashboards, catch faults, and win arguments about what actually happened last Tuesday. Then I started reading about world models, and it slowly dawned on me that we’d been sitting on something else entirely. Interleave the two streams and you get a long record of the world was like this, we did that, and here is what happened next. That record is exactly what a world model is trained on. We hadn’t been keeping logs. We’d been keeping lessons.