AI structural-accountability lab. Publishes a pre-validation framework that measures behavioral correlates of reasoning coherence, value coherence, and identity stability in AI systems under perturbation, without access to model weights or training data; inter-rater validation is the framework's stated next step. Frames alignment as representational continuity rather than behavioral compliance.
At a glance
- HQ: RTP
- Sector: DeepTech
- Team size: 1-10
See the full company record.
Founders, funding history and recent signals. Free accounts open immediately.
Sourced company notes 3 topics
Notes on Mentim · last checked Oct 1, 2026
Overview
Two-founder Triangle AI structural-accountability lab, also styled Mentim Research Group and Mentim Frontier Accountability Lab. Co-founded by Daphne Earley and Matthew Earley. Surfaced at the OpenClaw Triangle Meetup on 2026-05-28.
Framework
Publishes a pre-validation framework for measuring behavioral correlates of reasoning coherence, value coherence, and identity stability under perturbation. Its Mentim's Ontology represents observed material as State to Event to State transitions and defines five proposed primitives (Perceived Activation Density, Value-Coherence Balance, Transition Volatility Metric, Temporal Coherence Scalar, Narrative Role Encoding), combined in a proposed Structural Accountability Index. Designed to operate without model weights or training data. The public work does not establish that these scores predict correctness, safety, or operational failure; inter-rater validation is a stated next step.
Public artifacts
A 44-page V1.0 whitepaper and a Unified Annotation Schema.
Listed roles
All Triangle rolesChecking listed roles…