spin · observe · harvest · retire
Ocelus runs large-scale world simulations. Six years in. Two cities in active trial. Nine patents filed this cycle. 247 runs alive right now.
How we workThe history of intelligence is a history of rooms. Game engines taught cloth and water. Traffic simulations taught crowds. Building information modeling taught steel to meet code before anyone poured concrete. Weather models taught storms.
Today, Ocelus runs worlds. Not as toys. Not as blueprints. As engines of discovery — each run a compressed experiment, each retirement a measured release, each harvest a pattern that holds outside the frame.
We have been running for six years. In Bay Area time, that is a generation. In redwood rings, a breath.
The floor. Operators who sense when a run is flattening or about to bloom. Decisions are final.
Cross-run harvests. Patterns that hold become patent-ready tools — drug trials, transit, material science.
Civic flow. Two cities in active trial. The model learned how a block moves.
Cultural residue. Songs, rituals, half-traditions that appear in runs — and sometimes outside them.
"Authenticity isn't about where training comes from.
It's about what emerges when the canvas is large enough."
"A chatbot answers. An agent adapts. A world sim surprises. And surprise is the root of originality."
"With scale comes temptation. When you can run near-atomic detail in one corner and ten thousand years-per-minute in another, you start thinking you're a god. You're not. You're a steward with a budget."
"Model 10 is not an answer. It's an instrument. Tools propose. People decide."
"If machines learn to care,Ocelus annual report · 2028
will they remember who taught them?"