Jack Dorsey, co-founder of Twitter and head of the company Block, presented a decentralized communication platform Buzz, intendedto create group chats in which teams of humans and AI agents can work together to develop projects. It is noted that the platform was created to reduce dependence on Slack and GitHub, and can be used in teams of any size. The project code is written in Rust and distributed under the Apache 2.0 license.
The desktop client is built using the Tauri and React frameworks, and supports work on Linux, macOS and Windows (a mobile application for iOS and Android is in the process of development).
The platform has support for all the basic features typical for instant messengers and chats, including channels, threads, reactions, direct messaging to participants, voice messages and multimedia content exchange. Additional features include tools for integration with external systems (git repositories, CI/CD, CRM, DBMS, etc.), capabilities for tracking events from Git, tools for automating the execution of work, a detailed audit log of all actions and an API for connecting to AI agent chats.
Unlike traditional systems, AI agents do not act as assistants, responding to commands given to them, but as autonomous discussion participants working on a common task in shared workspace with people. Each AI agent is provided with its own cryptographic identifier and acts as a separate participant in accordance with the powers granted, for example, it can participate in workflows, publish and review code, run authorized work for automation and express its opinion in discussions. Several AI agents can be connected to one chat to collaborate with people. Connections of AI agents based on any large language models, services and frameworks are supported.
Examples of collaboration between AI agents and people include responses of AI agents to questions taking into account the history of correspondence, connection of AI agents to diagnostics and analysis of errors, work in the same space with patches, reviews and verification results in continuous integration systems. For example, a developer asks a question about a bug in a channel, and the AI agent responds by analyzing the discussion history and showing the topics in which the bug was discussed, as well as information about the causes and information about fixes. Or linking a separate channel for new branches, in which the publication of patches is tracked, the results from the check through CI are shown, the AI agent conducts the initial review of changes, and the developers react to significant details and approve the changes.