Clara Nexus: hypothesis, record, and registered predictions of a constitutional AI runtime

Clara Nexus is a developmental AI platform run as a business and studied as an experiment on the same live system. This paper sets out the whole position in one place: the hypothesis that an AI built on a control-theoretic account of intelligence can augment what a person does without weakening how they think; what has actually been shown so far and at what evidence level; the self-change register that exists, is gated by independent settlement and named human approval, and is still empty; and the preregistered experiment, identified by commit hash, by which the position can fail in public. It states plainly what is not yet shown, including that no human-outcome result exists, and lists the results that would refute it.