10 Levels of Jev: from Smart IFs to self-directing agents with system-one models
IndyDevDan · youtube · 2026-09-28
IndyDevDan walks through 10 levels of using TypeSafe's Jev — a JSON-programmed "system one model" for narrow decisions — alongside coding agents, with an open-source repo.
Core idea: keep the coding agent for hard work, delegate narrow decisions to a single-purpose model.
Levels cover: prompt-injection yes/no checks, multiple-choice triage, composite scoring of tickets/review risk, confidence gating for bash commands, model/agent/workflow routing, pre-tool guardrail hooks in a Pi coding agent, auto-compact decisions, dirt-cheap file reads without loading files into agent context, multi-file recursive glob questions, and the agent choosing its own Jev questions to classify test failures.
Takeaway: don't default to the biggest model — define the decision, pick the right tool, inspect confidence, validate on your production workload.
More from coding & agent
- NVIDIA Moves AI Agent Safety Controls Outside the Agent with OpenShell and BlueField-4 Sentry — shashib · 2026-09-29
- Swargs claims $50M+ transaction volume and 6,000+ listings on its AI agent marketplace — KyeGomezB · 2026-09-29
- Sonnet 5.5 effort settings make no difference in 15-task coding test: 9/15 at low, medium and high — every · 2026-09-29
- Claude Connectors: 7 essential integrations to stop copy-pasting between Claude and your apps — Aiden_Tech_Ai · 2026-09-29
- AI made building so cheap that 5 launches beat 5 months of thinking — alexmacgregor__ · 2026-09-29
- AI agent buys NBA tickets end-to-end, pivots on its own after seats sell out mid-checkout — armand_ruiz · 2026-09-29