Agentic web
How AI agents restructure discovery, value chains, and moats.
-
Multi-model councils as a governance primitive, not a demo
A credentialing decision is a governance decision, not a grading optimization. A multi-model council that adjudicates a credential verdict is the new registrar — a governance primitive (quorum, disagreement, conviction, appeal trail, versioning) that makes AI-issued credentials legitimate, not an eval demo for accuracy. Own the reference implementation.
-
The discovery layer breaks: agents don't browse your course catalog
Catalogs built for human browsing are invisible to AI agents. As agents mediate discovery, the move is credentials-as-evidence, not credentials-as-marketing — and whoever owns the agent-readable credential interface becomes the aggregation point.
-
Unbundling the degree: what agents will reassemble
The degree is a bundle — curriculum, signaling brand, accreditation, time, social-credit — shrink-wrapped into one artifact because verification was expensive. Agents make verification cheap, dissolve the bundle, and reassemble atomic, verifiable units at query time. Whoever issues the atoms agents compose wins.
-
Aggregation theory for credentials
Thompson's aggregation theory, applied to learning and credentialing — why credentials are becoming the new discovery layer, and what that means for builders of trust infrastructure.