-
Structured external note-taking for long-horizon agent state
Agents keep a persistent notes file (e.g. NOTES.md) outside the active context window to track long-running state -- progress, decisions, counters -- re-reading it as needed instead of relying on conversation history to hold that state indefinitely.
Why this verdict?
Demonstrated directly in Anthropic's 'Claude plays Pokemon' example, which tracked precise progress (e.g. exact level-ups) across thousands of steps via a notes file rather than in-context state, and recommended in Anthropic's context-engineering guidance as a standard long-horizon technique -- cheap, auditable, and it survives compaction and restarts in a way in-context state doesn't.