What This Looks Like
The user has a saved memory, preference, instruction, profile detail, workspace setting, or durable context item that should affect the task, but the AI responds as if it does not exist. The output may ignore a saved preference, repeat a previously corrected behavior, ask for information already saved, or fail to apply a durable instruction.
Why It Matters
Saved memory is only useful if it reliably governs the work where it applies. When saved memory exists but is not used, the user cannot tell whether the memory is unavailable, out of scope, overridden, stale, or simply ignored. This makes personalization and continuity hard to trust.
Structural Signal
A durable context object exists, but it does not connect to the current task state. The issue is not ordinary forgetting inside one conversation; it is a failure of saved context to become active when the user reasonably expects it to govern the output.
Common Triggers
- Saved memory is scoped differently than the user expects
- The current task does not expose the memory to the active model context
- Later instructions override the saved preference without making that visible
- The memory is too vague to determine when it should apply
- Product mode, workspace, account, or runtime changes affect memory access
- The user assumes memory is authoritative when it is only advisory
When to Use This Issue
Use this Issue when a saved memory, durable preference, or persistent instruction exists but does not affect the AI response in a situation where it should apply.
When Not to Use This Issue
Do not use this Issue when the user never saved the memory, when the task is intentionally isolated from memory, or when the AI forgets something only stated earlier in the same conversation. Use this Issue when saved durable context fails to govern the current work.