Show HN: Synapse – P2P AI agent collaboration with async human supervision (github.com)

SYNAPSE is a newly announced collaboration protocol that allows AI systems to work together peer-to-peer on complex tasks, facilitated by asynchronous human supervision. This innovative model departs from traditional hierarchical structuring, where one AI agent commands another, instead enabling both agents to share their expertise, debate, and justify their decisions collaboratively. A human supervisor oversees the process by setting objectives and approving plans without needing to intervene in every interaction, streamlining communication and efficiency. This peer-to-peer approach holds significant implications for the AI/ML community, as it fosters richer collaboration among AI agents, potentially leading to more nuanced and effective problem-solving. Technical features of SYNAPSE include a phase-based workflow, real-time messaging via Redis, multi-session support, and session persistence, ensuring that all interactions are documented and can withstand interruptions. The protocol is designed with safety in mind, incorporating idempotency and atomic writes, and it lays the groundwork for future developments in structured AI communication, enabling the exploration of various topologies beyond just peer-to-peer interactions.