Buzz: Share spare compute and run open models together over P2P networks (www.iroh.computer)

Block has introduced Buzz, a peer-to-peer (P2P) network platform designed to optimize AI model inference by allowing users to share spare computing resources effectively. With hardware costs soaring, Buzz addresses the need for more efficient resource utilization by enabling a community of users to collectively participate in running AI models through a decentralized framework powered by the iroh networking protocol. The platform integrates MeshLLM, facilitating opportunities for seamless inference without a central server, as members can opt-in to share their computing power while maintaining strict control over permissions and network access. The significance of Buzz lies in its potential to transform how AI development is approached in teams, shifting from isolated workspaces to collaborative networks where idle computing resources are leveraged for enhanced model performance and reduced costs. By eliminating the need for complex configurations typically associated with distributed systems, Buzz simplifies the process of connecting devices across varied networks. This innovative model not only addresses rising hardware prices but also fosters efficient collaboration among developers, paving the way for more ambitious AI applications using existing devices.