Start from your question
Choosing an AI to build on comes down to a few questions — pick yours, and go straight to the page that answers it.
One page for every reader — the role changes what leads, the depth changes how much shows. Nothing is held back: every section can show the rest in place.
What do you want to know?
"Which models should we adopt?" → Two ways to win for the lineup view, and the four measures for cost and quality. Results publish after the first scored round.
"How exactly is a model scored?" → How a model is scored, down to how to reproduce it.
"Can I trust the results?" → Why you can trust it and the Charter.
"What makes this benchmark different?" → Two ways to win and the real assignment.
"How do I help review the results?" → The human check.
"Is my lab's squad in the field?" → Who is in the bracket; labs with access can sign in.
"How can I see what each model actually did?" → What you can observe.