Sam Witteveen · 2026-09-29 · notable
Sam Witteveen — 'Using Jev In Your Agent Harness'
Sam Witteveen's 29 September 2026 video shows where the Jev decision model fits inside an agent loop, from model routing and risk gating to tool selection, with demos of skill disclosure and RAG re-ranking.
Sam Witteveen maps out where a fast decision model like Jev can replace full LLM calls in an agent.
What is it?
'Using Jev In Your Agent Harness' went up on Sam Witteveen's channel on 29 September 2026. The video returns to Jev, TypeSafe's model that answers typed questions instead of writing text, and looks at how to use it inside an agent harness.
How does it work?
The first half starts from the agent loop and the cost of full LLM calls, then treats Jev as a 'smart if statement' for model routing, risk gating, tool and skill selection, ranking and judging, and real-time triage. The second half wires Jev into the loop and runs two demos: progressive disclosure of skills and RAG re-ranking.
Why does it matter?
Many agent steps are small yes/no or pick-one choices that do not need a full language model call. The video also covers where Jev does not fit, including compound questions, prompt injection, and the choice between local and cloud models, which helps builders decide what to hand off.
Who is it for?
developers building LLM agents