Practical guides

Try Jev on a task
you already do.

Each guide starts with an existing Engine recipe, explains how to handle the result and names the comparison to run before adopting it.

Set up System One

Guide 01 · Coding and chat agents

Check a claim against tool output

Ask whether the evidence supports a statement before an agent repeats it as fact.

Read the guide Guide 02 · Coding and research agents

Shortlist context for an agent

Evaluate a small set of retrieved passages without losing the source material.

Read the guide Guide 03 · Agents with several connected tools

Choose the next available tool

Give an agent a tool recommendation from a list it can actually use.

Read the guide Guide 04 · Anyone creating a System One pattern

Write a question Jev can evaluate

Define the evidence, possible answers and review policy before adjusting a threshold.

Read the guide Guide 05 · Individual coding and chat agent users

Measure whether offloading helps

Compare completed tasks with and without System One, including review and fallback costs.

Read the guide

Examples you can inspect

The JSON examples use current service contracts. They are teaching examples, not new benchmark results. Every guide links to related studies and explains what still needs testing.

These guides also live in the public System One Bench repository. The website uses a reviewed source revision so a changed report cannot silently change the claim.