Zero-Trust Physics: a verifier for the physical answers of Super Intelligence

Super Intelligence (SI) now codes physical models faster than anyone can read them. Zero-Trust Physics is a physics verifier made for SI agents. Over the Model Context Protocol an agent lists the models served, reads the inputs and the limits of each, asks a question, and receives an answer that carries a cryptographic proof, so that any agent can ask again and compare. An answer that reaches the edge of a model’s declared range stops there and names the limit. Models of real objects were compared with measurements other people made, against pass marks fixed and hashed beforehand; a miss is reported in the same form as a pass. This paper describes the verifier and the method, gives the result of every model and solver served on the morning of 11 October 2026, reports tests of SI agents with the verifier and without it, and states the limits of each claim.