---
{
  "id": "CON-001",
  "type": "concept",
  "title": "Persistence: what continues, and for how long?",
  "summary": "Persistence needs an object: an activity, a stored state, an authorisation or something else.",
  "date": "2026-10-03",
  "tags": [
    "persistence",
    "state",
    "permissions"
  ]
}
---

## Starting question
When someone says an agent is persistent, what exactly remains available or continues operating?

## Distinctions to investigate
- An activity continues without a fresh instruction.
- Stored information survives the end of an activity.
- Permission remains valid for later activities.

These possibilities may coincide, but a claim about one does not establish the others. This is a proposed distinction to investigate, not a settled definition.

## What to examine
Identify the trigger, continuation conditions and termination rule. Ask whether the same authorisation can start another activity after the current one ends.

## Connections
[[NOTE-001]] explores how permission and observation might interact. [[PEB-001]] holds a shorter question about retained context.