Hey Jev team and fellow developers! I was halfway through writing up what I wanted to try with Jev when my access came through, so figured I’d say hi and

Hey Jev team and fellow developers! I was halfway through writing up what I wanted to try with Jev when my access came through, so figured I’d say hi and share! I’ve been building agent-harness, a small open-source project based on my own coding-agent setups, generalized so other developers can make it their own. The idea is to make how your agents work configurable: keeping context focused, delegating to appropriate agents and models, choosing how independently they operate, and deciding what counts as verified. Those preferences become switches backed by rules, skills, and hooks, with support for Claude Code and Codex. What makes me excited to try Jev is the combination of low latency and typed decisions with probabilities. These are small judgment calls I’d like the harness to make throughout a session, so the overhead matters. Getting probabilities back directly would let me set t