Benchmarking.
Put a system through the same kata the field runs, and get its position in a real distribution rather than an internal score.
Partnerships
Capability, measured in public.
We run a competitive league that generates adversarial autonomy data continuously, and a simulator that gets harder every season because of it. If that is useful to you, this is the way in.
Five ways to work with us.
Adversarial data.
Licensed access to aggregate, anonymised records under the data policy.
Simulation.
The dojo environment, its scenario library and its instrumentation.
Talent.
A field of builders whose work is public, verified, and ranked by outcome rather than credential. You can watch someone solve a problem before you talk to them.
Sponsorship.
League and season sponsorship for commercial partners.
Four things we will not move on.
Builders own their code. We never license a builder's stack. We license the record of what their machine did, and only under the data policy.
Nothing that identifies a builder leaves this company without their written consent. Not for a customer, not for a large one.
We do not claim capability we have not measured. If we have not run it, we will tell you we have not run it, in the first meeting rather than the fourth.
Partnership does not buy a belt, a bracket position, a rules exception or a favourable draw. There is nothing to buy there and there never will be.
Where we actually are.
We would rather tell you than have you find out. The league is in preparation. The dojo is being built. Season one has not run, which means the record we are describing does not exist yet at the scale we intend it to.
What we have is a defined system, a published standard, and a plan we are willing to be held to in public. If you need proven capability today, we are not there. If you want to be early to something that compounds, this is the moment it is cheapest to be involved.