Block launched Buzz on July 21, 2026, as a free, open-source workspace for people and AI agents. It runs on the Nostr protocol, is model-agnostic and decentralized, and is published under Apache 2.0. Each agent holds its own cryptographic key pair.
The workspace gives a team a shared place to inspect agent work. In a private chat, colleagues may need the user to forward the output and explain the context before they can build on it. A shared record can reduce that coordination work, provided it includes the information reviewers actually need.
What the team can see
Buzz includes channels, direct messages, voice, code, and workflows. Agents join channels with names, permissions, and activity streams. A code review, automation, or summary can appear alongside the team’s other work.
Jack Dorsey described it as “a new groupchat platform for teams of people and agents of all sizes.” That is a useful description of the interface: agent work appears in a space the team can use for coordination.
The identity design gives each message two signatures, one from the agent and another tying it to the authorizing human. That creates a record to inspect. Signed messages alone do not prove that every external action is logged or correctly authorized; a team still needs to check the tool permissions and action records around the workspace.
Decide what belongs in the shared record
Gallup reported on July 21, 2026 that 52% of US employees use AI at work, with rewriting text and searching among the main uses. That finding describes use. It does not establish how much happens privately or what records colleagues can access.
For your own team, examine a recurring AI-assisted task. Identify the context the agent used, the output it produced, the review decision, and any external action. Decide which parts colleagues need to see and which contain information that should stay restricted.
A persistent activity stream may help people avoid duplicate assignments and understand earlier decisions. The team also needs a way to find the relevant work without reading every message, and a policy for retaining or correcting old context.
Evaluate a shared workflow
Compare Buzz with the way your team currently shares AI-assisted work. Try a bounded assignment with a named reviewer and check whether another colleague can understand the result and continue the work from the record.
Inspect the authorization chain, permissions, and logs as part of that test. A shared workspace is useful when it makes the assignment and its review easier to follow. The team can establish that with a trial, without assuming that every private chat needs to move into a group channel.