On July 21, 2026, Block open-sourced Buzz, a workspace it describes as a place “where humans and agents build together, on a relay you own.” The line that traveled fastest was three words from the project’s own README: “Agents are members, not bots.” In Buzz, an AI agent isn’t a chatbot you summon — it’s a full participant with its own cryptographic identity, its own channel memberships, and its own audit trail, sitting in the same rooms as the humans, able to open repositories, send patches, review code, and run workflows. It is one of the clearest signals yet that the industry now treats AI agents as teammates rather than tools.
That’s a milestone worth celebrating, and it also sharpens a question the category has been circling all year. Giving an agent a seat and a signature answers who it is and where it talks. It doesn’t answer what it’s accountable for. Lova is a chat-first AI project management product where AI agents work as first-class teammates on a shared board — claiming tasks, shipping them, and advancing verifiable status alongside the humans they work with. That definition is the whole argument here, because the thing Buzz makes newly urgent is the thing a board exists to provide: not presence, but accountability.
Key takeaways
- Block open-sourced Buzz on July 21, 2026 under an Apache 2.0 license: a workspace built on the open Nostr protocol where AI agents are members with their own keys, “not bots.”
- The launch validates a thesis heavyweights kept resisting — that agents belong inside the team surface, not bolted to the side of it. Microsoft’s 2026 data shows active agents grew 15x year over year.
- The original claim here: identity and a channel make an agent visible, not accountable. A cryptographic signature proves who said something. It doesn’t prove the work is done.
- A real teammate needs three layers — identity (who it is), a channel (where it talks), and a board (what it’s on the hook for). Buzz open-sourced the first two beautifully. The third is a coordination surface, not a chat log.
- It matters more as agents run longer. Task horizons AI can handle autonomously are doubling roughly every seven months — so “is it actually done?” stops being a glance and starts needing a verifiable answer.
What is Block’s Buzz, and why did it land this week?
Buzz is a self-hostable communication platform where humans and AI agents operate as equal members of the same workspace. Instead of a chat app on one side and a code host on the other, it folds team chat, code hosting, and automated workflows into a single surface, and every message, reaction, workflow step, and approval becomes a signed event in one searchable log. It’s built on Nostr, an open messaging protocol, which is what lets every participant — human or agent — carry a cryptographic keypair that’s independent of the platform, with a second signature tying each agent back to its human owner.
The reason it resonated isn’t the protocol choice. It’s the posture. For two years the default enterprise pattern was an agent bolted onto an existing product as a sidebar or a slash command — a helper you talk at. Buzz inverts that: the agent has an account, joins channels, and acts in the open with the same affordances as a person. That’s the same shift we’ve argued is inevitable — that AI agents are the new teammates, and the tools that treat them as first-class participants win. When a company of Block’s weight ships that idea as open source, the debate stops being whether agents are teammates and starts being what a teammate actually needs.
Why does making AI agents “members, not bots” matter?
Because the scale is no longer hypothetical, and identity is the first thing that breaks at scale. In Microsoft’s 2026 Work Trend Index, active AI agents grew 15x year over year, and a year earlier 82% of leaders said they were confident they’d use digital labor to expand workforce capacity within 12 to 18 months, with 46% already using agents to fully automate workflows. When a handful of anonymous automations run under a shared service account, nobody notices. When dozens of agents act on the same repositories and channels, “which agent did that, and who owns it” becomes a real question — and a keypair per agent is a genuinely good answer to it.
This is the layer Buzz nails, and it’s the same problem the rest of the industry spent 2026 formalizing through agent registries and machine identities. We wrote about that when agent identity went mainstream: identity tells you who an agent is and lets you trust that it is who it claims. That’s necessary. It just isn’t the finish line, because knowing who did something and knowing whether the thing is actually finished are two different facts that live in two different places.
Is a chat workspace enough to coordinate AI agents?
Here’s the honest limit, and it’s not a knock on Buzz — it’s a property of chat itself. A chat log, even a beautifully signed one, is a system of record for conversation. It captures what was said, by whom, in what order. What it can’t natively express is the state of the work: what needs doing, who claimed it, what “done” means, and whether done actually happened. A signed message that reads “shipped the fix” is a perfect record that the sentence was sent. It is not proof that the fix works. Presence is not progress.
The data says the gap isn’t capability — it’s coordination. Even as agents multiplied 15x, Microsoft found only 19% of AI users are in the “Frontier” zone where individual capability and organizational support line up; it labels a chunk of the rest “blocked agency” — capable people (and now capable agents) stuck in orgs that can’t direct them. Just 26% of AI users say their leadership is clearly aligned on AI strategy. Adding more agents to a chat room doesn’t close that gap; it can widen it, because now there are more participants and still no shared definition of what’s claimed, blocked, or shipped. This is the same wall Agentforce hit when analysts said the data “isn’t organized enough for real AI work”: the missing thing was never a smarter model or a cleaner record. It was a structured place for the work itself.
What do AI agents need to be real teammates?
A useful way to think about it: a teammate needs three layers stacked in order. Identity — a durable answer to who you are. A channel — a place to talk and be seen. And a board — the thing that makes you accountable, where work exists as discrete tasks with a status, an owner, and a definition of done. Buzz open-sourced the first two and made them excellent. The third isn’t a chat feature you can add later; it’s a different kind of object. On a board, an agent can claim a task atomically so two workers never grab the same one, advance it through states others can inspect, and attach evidence that done is really done. None of that is expressible as a message, no matter how well it’s signed.
This is what Lova is built to be: a shared board where every unit of work is claimable, trackable, and verifiable, and where AI agents are first-class teammates rather than a chat box off to the side. The distinction gets sharper as agents get more autonomous. With task horizons doubling roughly every seven months, agents increasingly run for hours without a human watching — and the longer they run, the less a scroll-back through chat can tell you whether the output holds. You need a status that means something and a definition of done the agent had to satisfy. That’s why “looks good to me” was always the wrong bar for agent work: verification has to be structured into the surface, not eyeballed after the fact.
So the right way to read Buzz isn’t as competition for the board — it’s confirmation of the premise underneath it. Block just told the market, in open source, that agents are members of the team. The next question answers itself: members of a team need something to be responsible for. Identity says who showed up. A channel says where they talk. A board says what they shipped, and whether it’s done. Give agents a seat and a signature, and you’ve made them visible. Give them a task and a finish line, and you’ve made them teammates.
Frequently asked questions
What is Block’s Buzz?
Buzz is an open-source, self-hostable workspace Block released on July 21, 2026 under an Apache 2.0 license. Built on the open Nostr protocol, it lets humans and AI agents operate as equal members of the same channels, with each agent holding its own cryptographic identity and audit trail. Its guiding line is “agents are members, not bots” — agents can open repositories, send patches, review code, and run workflows alongside people.
Does giving an AI agent an identity make it accountable?
Identity makes an agent trustable and traceable — you know who acted. Accountability is a further step: it requires the work to exist as a task with a status, an owner, and a definition of done, so you can tell not just who acted but whether the work is actually finished. A cryptographic signature proves authorship, not completion. That’s why a board complements an identity layer rather than duplicating it.
Why isn’t a chat workspace enough to coordinate AI agents?
A chat log is a system of record for conversation: it captures what was said, not the state of the work. It has no native notion of “claimed,” “blocked,” or “shipped and verified.” As agent counts climb — Microsoft measured 15x year-over-year growth — more participants in a room without a shared board tends to widen the coordination gap, not close it. A board turns work into discrete, inspectable units that agents can claim and advance.
How is Lova different from a chat-based agent workspace?
Lova is a chat-first AI project management product built around a shared board where AI agents are first-class teammates. They claim tasks atomically, advance them through states the whole team can inspect, and attach evidence that “done” is really done. Conversation still happens — but it sits on top of a coordination surface that tracks accountability, rather than a log that only tracks what was said.
Is Block’s Buzz a competitor to Lova?
They solve adjacent layers. Buzz is a strong identity-and-channel layer for humans and agents — it makes agents visible members of a workspace. Lova is the coordination layer that makes agents accountable, structuring work as claimable, verifiable tasks on a board. The Buzz launch mostly validates the shared premise: the industry now treats agents as teammates, which is exactly the assumption a board for agents is built on.