Research notes

Does the agent understand the environment it's operating in?

Research on epistemic AI, hidden failures, and what agents know versus what they score.

Why embodied AI needs epistemic verification
Standard evaluations tell us whether an agent got the right result. For systems acting in changing physical environments, we also need to know whether they understood the world well enough to keep getting the right result.
June 2026
Focus areas

What we write about

Epistemic verification

Does the agent understand the environment well enough to keep passing as conditions change?

Hidden failures

Agents can succeed while relying on bad assumptions. We find what you don't yet know will break.

Agent redeployment

Turn agent redeployment into a repeatable system where every deployment teaches the next one.

Environment deltas

What changes between environments and which assumptions no longer hold.

Get an agent working once. Redeploy it everywhere.

Contact Pelagic
Release checkv1.8
Replay rundone
Compare behaviorreview
Block hidden failurerequired