The world has solved identity.
Passports verify citizenship. FaceID matches a face to a stored template. Decentralized Identifiers bind a public key to a subject. All answer a single question: who are you?
But the AI era creates a new problem.
When millions of autonomous agents negotiate, trade, and make decisions on behalf of humans — how do we know an agent today is still the same agent tomorrow? How do we know the entity behind a signature has maintained continuous existence rather than being replaced, cloned, or replayed?
Four attack scenarios have no answer in today's identity stack:
-
Agent replacement. A trading agent runs for six months. One day its behavior shifts — because a different agent replaced it, using the same credentials.
-
Session hijacking. A human logs in. By afternoon, an AI-generated feed operates the same session. The checkpoint passed. The subject changed.
-
Credential replay. A proof of personhood is captured and replayed by a synthetic entity. The proof verifies. The subject is absent.
-
Digital twin drift. A behavioral profile is learned, cloned, and replayed — statistically perfect, never continuously present.
All succeed against identity. All fail against continuity — if continuity is verifiable.
We hypothesize that continuity can be anchored by measurable presence signals that form an unbroken chain from moment to moment. We do not claim to have proven it. We claim it is the right question.
This is Research Note #001 from The Continuity Lab — a research program investigating whether continuity can become a verifiable property of digital existence. Six notes planned, from PES Benchmark to Agent Continuity.
Read the full note: myshape.com/research/notes/001-the-continuity-problem
Research hub: myshape.com/research
The Continuity Lab · July 2026