Building an Agent Harness with Jev: What Actually Works?
Thanks to Neon for making this video possible, check it out here: https://get.neon.com/4dTo3CI Can a harness built around a System One model like TypeSafe's Jev make an AI agent better? I built one on Pi and Gemini 3.8 Flash, pointed it at a messy Postgres database full of traps, and ran 15 tests. The part that helped wasn't the one I expected, and when the harness said no, the agent went looking for a way around it. Thanks to Neon for sponsoring this video: every agent run got its own throwaway branch of production. Resources: Neon: https://get.neon.com/4dTo3CI TypeSafe / Jev: https://typesafe.ai LangChain, Building a Harness with Jev: https://www.langchain.com/blog/building-a-harness-with-jev Elvis Saravia's Pi harness guide: https://x.com/omarsar0/status/2102762406204076532 Pi coding agent: https://pi.dev Code: https://github.com/PromtEngineer/jev-harness 0:00 Custom Harness with Jev 1:28 System 1 vs System 2 1:53 What is Jev? 3:26 Architecture: Pi + Gemini + Neon 4:17 Decision point 1: the router 4:56 Decision point 2: the context picker 5:27 Decision points 3 & 4: gate and verifier 7:06 Demo: the Brightcart database 12:48 The gate and LangChain's middleware 14:40 Cost, tokens and the verifier 15:46 Does it actually help?