Jack Dorsey is taking on Slack with Buzz, a group chat platform for teams and their AI agents

Jack Dorsey has spent the last decade building financial rails. Now he's building the conversation layer that sits on top of them. Buzz, announced Tuesday, is a group chat platform that treats AI agents as peers — not plugins, not webhooks, not afterthoughts. The desktop app runs on macOS, Windows, and Linux. The code is on GitHub. The license is open. The ambition is explicit: replace Slack and GitHub with a single workspace where humans and agents collaborate in real time.

This is not a sidebar integration. Buzz bakes agent participation into the chat primitive. An agent can be @-mentioned, assigned a task, handed a GitHub issue, and observed by the whole team in the same thread. The agent replies, commits code, opens pull requests, and surfaces context — all inside the conversation. No context switching. No relay through a separate dashboard. The agent is just another participant with a different latency profile.

Slack has spent years bolting on apps, workflows, and canvas views. It remains fundamentally a message bus with a marketplace. Buzz inverts the architecture: the conversation canvas is the workspace. GitHub projects live natively inside channels. A team can discuss a spec, spin up an agent to implement it, watch the commits land, and merge — without leaving the window. For founding teams that live in linear chat history, this reduces the tax of tool fragmentation.

The self-sovereign framing matters. Buzz instances are self-hosted. Source access means a team can add a custom agent runtime, a proprietary knowledge base, a compliance filter — whatever their workflow demands — without petitioning a vendor roadmap. That is the open-source lever Slack cannot pull. Enterprises that require data residency and audit trails get a path that doesn't involve sending proprietary context to a third-party LLM endpoint via Slack's API.

Georgios Konstantopoulos's Centaur validates the thesis from a different angle. Centaur runs inside Slack or via API, positioning itself as a "virtual employee." The fact that two credible builders converged on the same architectural bet — agents belong in the conversation, not adjacent to it — suggests the market structure is shifting. The question is not whether agents will inhabit chat. The question is who owns the conversation layer.

Dorsey's track record cuts both ways. Square and Cash App proved he can scale regulated, high-trust consumer products. Bluesky proved he can ship protocol-grade infrastructure that struggles to find product-market fit. Buzz sits closer to Bluesky in maturity. The desktop app is labeled "early stages." The agent framework is thin. The GitHub integration is a wrapper, not a deep fusion. Porting a production team onto this today would be reckless.

But the signal is clear. The next generation of startups — AI-native, agent-heavy, tool-skeptical — will not adopt Slack by default. They will evaluate whether the conversation layer can absorb their agent workflows natively. Buzz offers a reference implementation. Centaur offers a parasitic one. Slack offers a platform tax. The winner will be the layer that makes agents feel like colleagues rather than contractors.

Block's distribution advantage is real. Cash App and Afterpay give Dorsey a sales channel into millions of small businesses. If Buzz integrates with Block's financial rails — payroll, invoicing, capital access — the conversation layer becomes a revenue layer. That is a moat Slack cannot easily replicate. Microsoft Teams has the enterprise distribution. Discord has the community distribution. Block has the commerce distribution. Each will try to absorb the agent workflow.

The open-source strategy is also a talent strategy. Developers who want to shape their agent workspace will contribute to Buzz rather than build bespoke Slack apps. The compounding effect of those contributions — custom runtimes, domain-specific agents, compliance modules — could accelerate the platform past Slack's plugin velocity. But only if the core architecture survives contact with real teams.

Buzz will likely remain a curiosity for six months. The desktop-only distribution limits mobile-first teams. The agent SDK needs hardening. The GitHub sync needs conflict resolution. The search and history primitives need to handle agent-generated noise. These are not trivial gaps. They are the work of a year, not a quarter.

Yet the directional bet is sound. Work is becoming a conversation between humans and agents. The tool that makes that conversation fluid, auditable, and extensible will capture the founding teams of the next decade. Dorsey has planted a flag. Whether Buzz becomes the standard or merely inspires the standard is secondary. The conversation layer has been contested.