Sadiya Bhawania
Full-Stack · AI Systems · LLMs

/ RESEARCH ILLUSTRATION

Words are not the world

Most AI today is trained to guess the next word. A world model is trained to guess what happens next in the real world. These are not the same thing. Scroll to see the difference, and why it is worth a billion dollars of research.

Two kinds of guessing

Language model

Theballrollsoffthetable

Predicts the next word from patterns in text.

World model

next

Predicts the next state of the world, as a compressed idea rather than exact pixels.

The loop inside a world model

This four step loop is roughly what the research toolkit (stable-worldmodel) standardizes, so different labs can be compared fairly. Watch it cycle.

CameraEncoderPredictorPlanner

A frame comes in from the world.

Then it repeats. The planner acts, the world changes, a new frame comes in, and the loop runs again.

Try it yourself

This is a language model pretending to be a world model. It is not a real one. I built this first to understand the idea myself, and to demo it for anyone curious about what comes next, beyond LLMs.

Pick a scene

Then do something

Choose an action to see what the model predicts.