Products
Research
How we test AI infrastructure, written up so others can check our methods.
How tokenizer differentials, relative error distributions, and behavioral anchors expose surrogate LLMs.