Skip to content

2089-block-buzz-agent-workspace

Directory actions

More options

Directory actions

More options

Latest commit

 

History

History

2089-block-buzz-agent-workspace

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
README.md
README.md
 
 

README.md

topic agents
type market-research
status research-complete
last-validated 2026-07-27
related-docs 601, 2030, 2088
original-query https://github.com/block/buzz - research this GitHub repo (Block Inc's buzz project). What it is, what it does, the stack, license, activity/maturity, and any relevance/applicability to the ZAO ecosystem (agents, music, dev-workflows, infra). Tier STANDARD.
tier STANDARD

2089 - Block's Buzz: humans + AI agents as peers on a Nostr relay

Goal: Assess Block Inc's open-source "Buzz" workspace and decide whether ZAO should adopt it, borrow from it, or just monitor it - specifically for the multi-agent + human coordination problem ZOE already has.

Key Decisions (recommendations first)

Decision Call Why
Adopt Buzz as ZAO's workspace NO Adoption is a migration off Discord + GitHub + Telegram, not an integration. ZAO's stack is Vercel/Supabase/Neynar (web2 SaaS + web3 auth); Buzz is Nostr-native. Wrong cost/benefit today.
Borrow ONE idea: signed, receipted cross-agent handoffs YES - investigate Buzz's core novelty is that every agent action (message, PR, review, merge) is a signed Nostr event with a cryptographic audit trail. ZAO's cross-lane coordination is currently unsigned manual clipboard hand-offs (this very session: ZOE lane <-> ZAOcowork lane via copy-paste). That is the exact gap Buzz's model closes. This also rhymes with the DreamNet receipt work (doc 2030).
Monitor Buzz as a reference architecture YES 14.6k stars in ~4.5 months, Block runs it internally to replace Slack + GitHub. It is the most credible real-world "agents as team members" design to watch for agent-scaling + CLI-agent patterns.

One-line take: Buzz is a strategic learn, not a roadmap item. The idea worth stealing is signed/receipted agent handoffs; the product is not for ZAO's music/community vertical right now.

What it is

Buzz (github.com/block/buzz, Apache-2.0, launched 2026-07-21) is Block Inc's open-source workspace that unifies team chat, code hosting, and AI-agent coordination on a single Nostr relay per community. Humans and AI agents are peer participants - each with a cryptographic identity - and every action is a signed event, giving a tamper-evident audit trail. Block is dogfooding it internally to replace both Slack and Chat and GitHub. It is self-hostable (an org runs its own relay, keeping data off Block's servers).

Findings

Dimension Detail
Core capability Unified messaging (channels/threads/DMs, media comments anchored to code frames) + git (patches, CI results, reviews, merges stored as signed Nostr events) + YAML workflow automation with approval gates. Agents get repo access, issue triage, workflow orchestration, channel creation - as full members, not bot integrations.
Agent interface CLI-first, supports tool-calling LLMs - Goose (Block's own), Codex, and Claude Code named explicitly.
Stack Rust 54% (13.5M bytes, the relay backend, Axum + PostgreSQL + Redis pub/sub + S3-compatible storage) - TypeScript 39% (frontend/tooling, React) - plus Dart/Kotlin/Swift mobile SDKs.
License Apache License 2.0 (permissive - forkable).
Maturity 14,571 stars, 1,242 forks, 801 open issues. Created 2026-03-06, last push 2026-07-27 (same day - extremely active). ~4.5 months old, production-grade (Block runs it internally).
Community reception HN launch thread ~374 upvotes / ~331 comments. Serious technical discussion; skepticism centered on "agent choreography" complexity.

Specific numbers: 14,571 stars; 1,242 forks; 801 open issues; created 2026-03-06; Rust 54% / TS 39% of the codebase.

Relevance to ZAO

Where it lands (yes):

  • The coordination problem is ZAO's problem. This session ran two Claude lanes (ZOE + ZAOcowork) coordinating over the same Supabase board via manually copy-pasted clipboard blocks - with an explicit "do not double-run the SQL" caveat because there was no shared, verifiable state. Buzz's signed-event model is a principled answer to exactly that: agents as peers writing to one auditable log. See the fleet + orchestration surfaces (doc 601 agent-stack cleanup).
  • Signed agent work aligns with ZAO's verifiable-identity bias (Farcaster signers, XMTP identities) and with the DreamNet receipt emission ZAO already does (dreamnet.receipt.v1, doc 2030). The "every action is a receipt" idea is not foreign - Buzz just makes it the substrate.
  • Same LLM tier. Buzz's CLI targets Claude Code, which is ZOE's own coder brain - so the agent-interface patterns are directly readable.

Where it does not (no / unclear):

  • Migration, not integration. ZAO lives on Discord (cowork), GitHub (code), Telegram (ZOE). Buzz would replace all three. Not worth it now.
  • Different agent model. Buzz = agents as channel peers; ZOE = orchestrator-centric (one Telegram bot, DMs for questions, group for status). Different UX + governance.
  • Vertical mismatch. Buzz is generalist team tooling; ZAO is vertical for music creators + web3 communities (SongJam, ZABAL Games, COC Concertz). No creator-platform angle in Buzz.

Also See

  • Doc 601 - the ZAO agent-stack cleanup (ZOE as the one orchestrator)
  • Doc 2088 - ZAI Discord capture (another "borrow OSS, don't adopt wholesale" call)

Next Actions

Action Owner Type By When
Read this doc + decide: monitor-only, or spike a signed/receipted cross-lane handoff for ZOE (vs the current manual clipboard) zaal Decision 2026-08-05
If spiking: draft a design doc for signed ZOE<->cowork handoffs, reusing the existing dreamnet.receipt.v1 emission (doc 2030) rather than adopting Nostr zaal Research doc 2026-08-12
Re-validate Buzz's maturity + any ZAO-relevant features (staleness check on this fast-moving repo) zaal Doc update 2026-09-27

Sources