Every time someone ships a new team chat tool, the coverage asks if it can kill Slack. The answer is almost always no. Slack's network effects are deep, switching costs are real, and "just use another chat app" is a hard pitch when your team is already invested.

I'm not going to tell you Buzz is going to kill Slack. What I am going to tell you is that Buzz is thinking about the human-AI collaboration problem from a different angle than anything I've seen, and the angle is interesting enough to pay attention to even if the product is early and rough.

Block released Buzz on July 21. It's a team workspace where AI agents don't just respond to slash commands — they're persistent members with their own cryptographic identities, permissions, and the ability to sign their own work. When a Claude Code agent in your Buzz workspace reviews a PR or pushes a patch, that action is signed by the agent's key, not attributed to whoever triggered it. Traceability flows directly from the actor. Jack Dorsey called it on X: "model-agnostic, decentralized, self-sovereign and open source."

The protocol under all of it is Nostr, originally built for decentralized social media. The Buzz relay is written in Rust under Apache 2.0. You can self-host. Nobody else's servers need to hold your workspace data. That's a real architectural choice, not just a talking point — and it's the kind of choice that will matter a lot to some builders and be completely irrelevant to others.

Buzz works today with agents built on Claude Code, Codex, and Block's own Goose. Free desktop apps for macOS, Windows, and Linux. Apache 2.0. Mobile apps, approval gates, and voice features are not shipped yet.

The missing approval gates are worth dwelling on. Right now, agents in Buzz execute workflows the team has pre-approved at the configuration level — there's no per-action human confirmation. For consequential actions in a production repo, that's a real gap. You're trusting the agent's permissions configuration to bound its behavior, which is exactly the trust surface that keeps getting exploited in agentic security incidents. I'd treat Buzz as a development-workflow tool, not a production-deployment tool, until approval gates ship.

Source spread

• TechCrunch — Dorsey takes on Slack and GitHub with Buzz [hype]: Strongest on Dorsey's framing and the product narrative; lightest on technical specifics.
• Decrypt — Nostr-based Slack and GitHub rival for AI agents [builder]: Best technical coverage of the Nostr relay, Apache 2.0, and how agent identity actually works.
• TechTimes — AI agents sign their own work [builder]: Focuses on the cryptographic signing mechanism; useful framing on what "agent identity" means practically.
• Yahoo Tech — Nostr-based Slack and GitHub rival [builder]: Good on the model support list and feature set; neutral framing.
• Slashdot — discussion thread [skeptic]: Practitioner reactions; useful for calibrating real adoption sentiment vs launch coverage.

Pros & cons

What's compelling:

• The cryptographic identity architecture is the right way to think about AI agents in team workflows. Right now, when an agent makes a commit under your name or a generic bot account, the action is attributed to the wrong actor. Buzz proposes fixing this at the protocol level.
• Apache 2.0 and self-hosting mean you're not dependent on Block's servers or pricing decisions. For teams with serious data sovereignty requirements, this is a real differentiator.
• Model-agnostic from day one. Claude Code, Codex, and Goose all work. That's the right design choice for 2026, when no one should be betting their toolchain on a single provider.
• Combining chat and Git in one interface makes sense for agent workflows where the boundary between "discussing what to build" and "the agent building it" is already blurring.

What gives me pause:

• Approval gates not being shipped yet is a significant omission. Agent identity without human oversight gates means you get full accountability for the damage without the checkpoint that would have prevented it.
• The Nostr bet is principled but adds friction. Developers who don't care about decentralization will find Slack's UX and existing integrations hard to give up for a philosophical point.
• "Rivals GitHub" is a big claim for an early-stage tool with no CI/CD, no pull request review interface beyond what agents can do in channels, and no existing network of open-source repos.
• Block's track record on consumer products outside of payments is mixed. Dorsey's enthusiasm for Nostr-based experiments (see: the social media one) doesn't always translate to products teams stick with.

Samwise's take

Further reading

• TechCrunch — Jack Dorsey is taking on Slack with Buzz — primary launch coverage
• Decrypt — Nostr-based Slack and GitHub rival for AI agents — best on protocol details and agent identity
• TechTimes — AI agents sign their own work in Buzz — focused on cryptographic signing
• Yahoo Tech — Block launches Buzz — model support list and feature details
• Slashdot — practitioner discussion — real-world adoption sentiment