Sensory reality engine
MASH
Multi-Agent Semantic Hallucination. It does not model a world from the outside. It puts you inside one and renders it as you move.
What it does
MASH weaves a continuous sensory field from multiple agents working the same scene at once. Each contributes a strand: what is visible, what is heard, what the space is doing at the edges of attention. The result is not a description handed to you after the fact. It is a place that holds together while you are in it.
The word hallucination is exact rather than pejorative. The engine is generating perception, and the interesting problem is coherence: keeping the strands agreeing with each other frame to frame, so the world does not dissolve when you turn around.
What it looks like from inside
Rendered by MASH
Not concept art. This is a frame the engine produced inside a running world, captured through its own snapshot system, greeting a visitor in the world's own voice.
The loom decided the composition, the palette and the lighting. Nobody art-directed it.
The visual loom
The loom is what makes it Maisie's rather than anyone else's. Strands of image, motion and light are woven under tension, each agent pulling its own thread, and the weave is what you see. Composition, palette and lighting are parameters of the loom, not requests made of it.
This is why MASH sits in the art register. The output is not a report about a world. It is the world's surface, made continuously.
How it differs from MASS
MASS, the Multi-Agent Semantic Simulator, models a system from the outside and hands you an executable artifact about how it behaves. MASH places you inside as a participant. They share a foundation and answer different questions, and anyone who finds both repositories should be told which is which.