Preview in 5.0.x

What gets kept,
stays visible.

Per-message visibility badges (will this enter long-term memory? whose memory?), an end-of-output prompt to pin or bookmark, a per-message override menu, and three archetype presets in settings — “remember everything,” “ask each time,” “remember only what I pin.” Asymmetric: the user sees more than the model does about what's being kept.

Next patch


The principle

The user is in control of memory.

The platform's memory system is a real asset — the more it remembers about your work, the better it gets at it. But that asset only works if the customer trusts the boundary of what gets kept. The next 5.0.x patch makes the boundary visible at every step instead of hiding it behind a single global toggle.

The design principle: the customer sees more than the model does. The model never reads what the customer sees about retention — that's UI, not context. The customer gets a clear, persistent view of what's heading into long-term memory, where it's heading, and how to override it.


What ships

Four surfaces.


What's pending

Honest gates on this page.

Still in flight

  • Per-message badge in the chat shell
  • End-of-output pin prompt (across web, macOS, iOS)
  • Per-message override menu
  • Three archetype presets surfaced in settings
  • Memory-scope dependency: the badges need memory scoping to surface the four scope tiers
  • Tenant-default preset (admin sets the starting posture for new users)

This page updates as each piece lands. The release notes are the formal cut.


Read next.

For the capability side of the same UI work, see capability-driven self-awareness UI. For the memory model these badges describe, see memory scoping. For the full 5.0.x roadmap, see what's next in 5.0.x.