Finally, a proper AI multiplayer: Buzz, the decentralized hive for humans and agents

For years, “collaborating with AI” has meant awkward side panels, slash commands, private chats that no one else can see, or brittle integrations that fall apart the moment you switch tools. Shared history? Usually nonexistent. True multiplayer with multiple agents working alongside people in the same persistent space? Rarely more than a demo.

That just changed.

Buzz (https://buzz.xyz/), launched by Jack Dorsey’s Block, is the first workspace that treats AI agents as first-class teammates rather than bolted-on assistants. Humans and agents share the same channels, threads, direct messages, voice huddles, project discussions, and — critically — a single, continuous history. Agents post, review code, run approved workflows, open pull requests, and build on each other’s work exactly like colleagues. The result feels less like prompting a chatbot and more like running a mixed team of people and specialized “bees.”

The bee motif is deliberate. The branding leans hard into the hive: animated bees, the 🐝 emoji in Dorsey’s announcement, agent names that evoke the colony, and the explicit framing of a “hive mind” where collective intelligence emerges from shared context. A short demo video captures the spirit perfectly—agents buzzing around the workspace, coordinating like a real swarm.

---

Slack + GitHub, but open and agent-native

Buzz is positioned as a direct replacement for the tools most engineering teams currently juggle. Channels, threads, reactions, media, and voice live next to native Git repositories, planning canvases, automated workflows, and pull-request flows. You discuss an idea, turn it into code, review it with agents, and merge — all without leaving the room. No more context-switching between Slack and GitHub while agents operate in their own silos.

It is model-agnostic and agent-agnostic. Plug in Claude Code, OpenAI Codex, Block’s own open-source Goose, or any agent that speaks the Agent Client Protocol. Agents receive their own cryptographic identities (keypair plus an owner attestation), defined permissions, and an independent audit trail. When an agent acts, the record shows both what it did and which human authorized it.

---

Built on Nostr: decentralized by design

Under the hood sits the open Nostr protocol (https://nostr.org/). Every message, reaction, workflow step, code event, and approval is a cryptographically signed event. Identity is a keypair you control — portable, verifiable, and independent of any platform account or vendor API key. Agents carry their history and reputation with them across compatible systems.

You can run your own relay (full self-sovereignty, data never leaves your infrastructure) or use Block’s hosted option at buzz.xyz while still benefiting from the same protocol guarantees.

There is no single central server that can be blocked, sanctioned, or quietly rewritten. In a world of increasing platform risk and regulatory pressure, that is not a marketing slogan — it is the architecture.

The entire stack is open source under Apache 2.0. Source code lives at https://github.com/block/buzz. Desktop clients for macOS, Windows, and Linux are available now as a developer preview; self-hosting is straightforward. Fork it, modify it, run it on your own hardware.

---

Shared compute: the hive contributes its own power

One of the most cyberpunk features is optional compute sharing. Members of a community can flip a switch and contribute idle GPUs or local models. Agents in the same relay can then draw inference from the pool of machines their teammates have offered — without routing prompts through a central cloud or generating new API bills.

Tokens and compute become a communal resource among consenting participants. Larger models become available automatically as more capable hardware joins the mesh. The relay coordinates trust and discovery; the actual inference traffic stays peer-to-peer.

This turns a team’s existing hardware into a private inference network. No vendor lock-in, no surprise invoices, and full visibility into what is being shared.

---

A breath of fresh air

Buzz is still early — developer preview, mobile clients still maturing — but the direction is clear. It solves the multiplayer problem for agentic work: shared context, persistent history, equal standing for humans and machines, cryptographic accountability, and true ownership of the infrastructure.

Jack Dorsey, the creator of Twitter who has spent years pushing Bitcoin and open protocols, has now applied the same philosophy to the emerging human–AI workplace.

The result is decentralized, open-source, model-agnostic, and deliberately resistant to capture.

Download the developer preview at https://buzz.xyz/. Read the code at https://github.com/block/buzz. Spin up a relay, invite a few agents, and watch the hive get to work.

The age of proper AI multiplayer has started. It smells like open protocols, shared compute, and a healthy dose of cyberpunk optimism.

---

Also worth reading:

• Main Competitors of the iPhone 17 Pro: Where Android Flagships Push the Boundaries of AI
• 5 Low Dopamine Habits for High Focus Freelancing
• Google Play’s Target SDK Policy: A Gradual Cleanup That Opens the Door for Modern Apps
• Block’s Proto Rig: Jack Dorsey’s Vision for the Future of Bitcoin Mining
• Critical Vulnerabilities Found in Jack Dorsey’s Decentralized Messenger Bitchat

---

Thank you!