The research group

Mentim was founded to address a critical gap: output-level evaluation cannot measure the structural condition of autonomous systems. While industry standards focus on behavioral output, we focus on structural coherence — how a system holds together while it works, not just what it produces.

Our work rests on two published foundations. The Mentim Ontology defines what is measured: five structural primitives and the composite Structural Accountability Index (SAI), a trajectory-sensitive metric of internal coherence. The Unified Annotation Schema defines how observations are recorded, so measurements remain comparable across systems and studies. Together they provide a standardized language for measuring how systems maintain stability under pressure and context shift. Both are open access on the Science page.

Manifest Prime is the first product built on that research — runtime structural measurement for autonomous multi-agent systems, with preserved, replayable evidence.

Mentim engages with researchers, laboratories, and institutions to advance structural accountability in autonomous systems. For autonomy programs, we operate as a measurement work package on larger teams — see working with programs.