Conol 0.41.1 — Shared memory that's truly shared

A small release that closes the gap between how things looked and how they actually worked: memory saved in a shared workspace now belongs to the workspace, and the billing page tells you exactly what a scheduled plan change will do.

Workspace memory is shared with every member

Why. Memory is how your agent gets smarter over time — preferences, project context, decisions it should remember. In a shared workspace that knowledge is supposed to be a team asset: whoever asks, the agent should recall what any member taught it. Until now, workspace memory was quietly scoped to whoever created it — you'd see the entries written from your own chats, but not your teammates', and they wouldn't see yours. Each member was effectively talking to a different agent.

How to use it. Nothing to set up. Memory the agent saves with workspace scope is now visible to — and usable by — every member of that workspace, including entries created before this release. You can browse it under Skills & Memory → Memory, where each file shows its scope; workspace-scoped files are the shared ones. Your global memory (personal preferences that follow you across workspaces) and per-chat session memory stay private to you, exactly as before.

Fixes

• Scheduled downgrades name the plan — if you schedule a plan change for your next billing cycle, the billing page now says which plan you're moving to and when it takes effect, and that plan's card carries a "Scheduled" badge instead of leaving you guessing what was queued.
• The composer always shows your chat's model — some chats created on the GPT-5.6 models during launch week could show an empty "Select model" picker in the message box. They now display (and keep using) the model they were started with.
• Sub-agent bars show the model name — when your agent spawns sub-agents, the little bar for each one now labels it with the model it runs, cleaner than the technical identifier it used to show.