An AI research lab · trusted by OpsCom
We research agents, models and the infrastructure they run on, keep the ideas that spark joy, and ship them as tools people can rely on.
From orbit to the bench
Scattered experiments become tools you can keep.
Trusted by
OpsComFrom the lab
Work first, play after.
A strict macOS task enforcer. granny asks what you are doing today, blocks the entertainment while you work, checks your claims, and unlocks the fun only when the list is done. She keeps a ledger of your day offs and is not impressed by them.
Neural dots that learn to hunt.
An artificial-life simulation where every dot runs its own tiny neural network. Three species share a field in a rock-paper-scissors loop, and a genetic algorithm breeds the best hunters. Nobody programs the behavior; it comes out of local sensing and selection pressure.
What the lab does
01
We chase ideas that might be nothing: agent architectures, fine-tuning runs, odd little interfaces. Most stop here, cheaply, and that is the point.
02
The ideas that survive get hardened into tools people can depend on: documented, tested, maintained, and free wherever we can make that work.
Research
No fixed roadmap. These are the areas the lab keeps coming back to, in prototypes and in products.
How we work
01
A rough, running thing in days, not months. The point is to feel whether the idea has legs.
02
Most experiments stop here. The ones that make us grin get written down and scoped into a product.
03
Tests, docs, edge cases — the boring work that turns a demo into something people can trust.
04
We release early, listen, and maintain what we ship. No abandoned tabs, no forgotten repos.
Tell us what you are making — or what should exist. Someone from the lab replies within two business days.