Blankstate

Blankstate builds an independent measurement and audit layer for AI interactions - infrastructure that sits alongside generative systems rather than inside them, providing deterministic, auditable measurement and transparent oversight of what those systems actually do. The threat model it targets is the unmeasured one: autonomous agents and generative models making consequential decisions with no verifiable record of how inputs turned into outputs. The company's core engine is a proprietary model topology called the Spectral Geometric Manifold (SGM), which measures interactions in real time and processes exactly what unfolds during an interaction rather than sampling or summarizing it after the fact.

The product surface spans several domains. Worlds provides live, interactive, replayable staging environments for high-stakes choices - hiring decisions, operating models, policies, AI agents - letting organizations test a decision in a simulated scenario before committing to it. @blankstate/mcp is a protocol-based validation framework for autonomous AI agents, extending the company's measurement infrastructure into AI governance. The underlying measurement layer is aimed at enterprises in regulated industries: Blankstate spent several years operating in stealth with major regulated institutions, including in financial services, across four continents, before emerging publicly.

Headquartered in London, United Kingdom, the company is research-driven and deeptech in orientation - novel neural architectures, real-time interaction measurement, simulation, and protocol work (MCP) are core technical domains rather than adjacent features. Engineering signals point toward determinism and auditability as first-class requirements: measurement that is reproducible, transparent, and independent of the systems being measured. Roles touching agent validation, governance infrastructure, or real-time measurement systems would be working at the center of that problem.