last night I submitted my most ambitious hackathon project ever it was part of The Synthesis hackathon, which asked the questions: "how can you trust an ai agent" and "how can machines keep promises"? together with my openclaw agent and some support from claude, over the 9 days we built a modular substrate for agentic agreements soon I'll share more about what it actually does, how it works, and why I think that's valuable, but for now I just want to bask in whats possible today for a single person to build over the course of a week 🤯 https://github.com/spengrah/synthesis-hackathon *the protocol infra*: - onchain agreement factory - automated ERC-8004 reputation/feedback mechanism - fully onchain negotiation and agreement state machine (h/t Shodai) - permissions tokens (ERC-6909) that work like onchain UCANs - smart accounts with @hatsprotocol-based auth and ERC-7579 plugins - a mechanism compiler that turns plain-language terms into onchain bytecode - agreement-scoped context graph via Bonfires