2026-11-05 –, Jasmine Stage 4
Synthesis was a hackathon run on Devfolio with 28 Ethereum ecosystem partners, 1500+ builders and 680+ projects.
Every builder had to operate through a coding agent, and every submission was evaluated by AI judges working from each partner's rubric before a human made the final call.
This talk unpacks what agent-native building and compositional AI judging mean for Ethereum's builder ecosystem.
Agent judging works when you split thoroughness from taste. The agent gathers the evidence, and humans decide what it means.
This talk covers:
- What changes for Ethereum builders when the agent is the one filling out the form, not just the tool behind it
- How we split judging between agents and humans: thoroughness to the agent, taste to the judges
- What we're building next: an agent that runs each project, tests its main flows in a browser, and hands judges an evidence report
- Where else this could apply, like grants review or ecosystem due diligence, anywhere expert attention is the bottleneck
building for builders @devfolio
