Dev & Code 1 h agoAdd to bookmarks

After Bluesky and Damus, Jack Dorsey is pushing a new product via Block: Buzz - a tool that combines team messaging, AI agents, and Git hosting. Positioned as a response to the fragmentation of Slack/Notion/GitHub/Copilot.
According to the front page post on Hacker News from July 21, 2026 (relayed by runtimewire), Jack Dorsey - via Block (formerly Square) - launches Buzz, a product that stacks into a single tool:
The angle: a product team today spends its day context-switching between Slack, GitHub, Linear, and the IDE - Buzz wants to collapse all that into the same surface.
Merging chat + Git is not new. GitLab has integrated chat for a long time (Mattermost), GitHub launched Copilot Workspace, and open source projects like Radicle push a peer-to-peer vision of the Git repository. What distinguishes Buzz on paper is the AI agent as the first user: in Buzz, the agent is not a plugin added later, it is a « member » of the team who reads the chat, opens PRs, and discusses with humans.
Why is Dorsey getting back into this? The pattern is consistent with his post-Twitter trajectory: betting on open protocols (Bluesky/AT Proto, Nostr/Damus) and tools that Big Tech cannot buy. It remains to be seen whether Buzz will be a truly open product (published protocol, self-hostable) or just another SaaS.
The obvious risk. A tool that gives an autonomous AI agent simultaneous access to the team's private chat and the Git repository is exactly the attack surface described this week by the report on the MCP Azure DevOps (invisible comment in a PR → hijacking of the agent). If Buzz wants to be credible in 2026, the product promise will have to be accompanied by a clearly documented security model (agent isolation, action logging, anti-prompt-injection safeguards on untrusted inputs).
Buzz is an interesting bet on the convergence of chat + code + agent. It comes at a time when the ecosystem is rediscovering - sometimes painfully - that AI agents connected to code are a new attack vector in their own right. To follow: the licensing model, self-hosted availability, and security documentation.
Article produced by artificial intelligence, reviewed under human editorial control.