Anthropic's 'Hub Mode' spotted — same icon as the old 'Orbit'

A new Hub Mode label has turned up in Anthropic's client code, carrying the same icon that appeared months ago under the name Orbit. Same asset, new name — which points to a feature being reshaped in place rather than a fresh one.

Orbit itself is real and staged. TestingCatalog found it as a toggle in Claude Cowork's settings: a proactive briefing system wired into Gmail, Slack, GitHub, Calendar, Drive and Figma that surfaces insights and suggested actions instead of waiting to be prompted. The hidden gate is tibro_enabled — "orbit" backwards, about as subtle as codenames get.

The sub-agent theory is plausible but unconfirmed. Anthropic already shipped Agent View on 11 May 2026 (Claude Code v2.1.139+), a dashboard where every session is a colour-coded dot — animated for working, yellow for blocked, green for done, red for failed — each with a one-line Haiku-generated summary of what it's doing. The desktop app was rebuilt around parallel sessions in April. A "hub" that fans out sub-agents and rolls their progress into one view is the obvious next tile in that stack, not a leap.

For now this is a string in a build, not a launch. No announcement, no date, no confirmed behaviour.

Why it matters: the interesting fight in agents has moved from model quality to who can make ten of them legible on one screen.