Label everything unproven
Why every claim on our site is either measured on real infrastructure or marked pending.
Security marketing has a habit of describing the product it plans to build. We decided early to do the opposite, and to make it visible.
The rule
Every capability we claim traces to a real run on real infrastructure. Every figure we have not measured on hardware is marked pending. On this site that shows up as two labels, and we use them literally.
Kubernetes: ~5.5 s eviction on zero-day scenarios, 100% prevention, 20 / 20 stateful jobs preserved, zero false positives. Real cluster, n=20, reproducible.
AI inference: design-complete, mechanisms built and CI-tested. Effectiveness figures wait for real GPU runs.
Why bother
Because the people who buy this are the people who get asked hard questions afterwards. A number that turns out to come from a slide, not a cluster, costs more trust than it ever earned.
It also keeps us honest internally. A capability is not done when the code merges. It is done when it has run somewhere real and produced a number we are willing to print.
That discipline is the product.
What is next
The next label we want to flip is the GPU one. If you run an inference fleet and want to help produce the first measured results, we are looking for validation partners, and your stack would shape the test matrix.
Get the next one when it lands.
No schedule. We write when there is something proven to say.