The forms that survive transformation.
An eigenform is what a transformation cannot change — the shape that comes back as itself. We take the name as a research posture: beneath every learned representation there are invariants doing the real work, and much of what goes wrong in AI systems today is someone mistaking the coordinates for the form.
Our first line of work is measurement validity for AI systems — whether the scores agent systems act on (similarity, drift, relevance) measure the constructs they are deployed to enforce. The instruments ship open; every number reproduces offline from frozen evidence.
- polaritycheck — corpus, harness, and frozen results for a construct-validity audit of embedding-cosine quality gates in agent systems. Paper in submission.
Held loosely, stated plainly: past vector space, toward the symmetries that bind representations across scale — which structure survives when everything else is transformed away.
Founded by Scott E. Frias — Creator/Founder, Eigenforma; Co-Founder, Freemind Labs.
