JevMade field notes / Calibration procedure

A guide to testing an AI tool that chooses from options

Learn how to test an AI tool that picks from options. This guide explains how to use practice examples to improve the instructions you give the software, ensuring it makes reliable choices on real tasks.

Original by smkrvEvaluation

Listen to this guide

JevMade’s plain-English explanation

0:00 /

AI narration

Credits and license

“jev-calibrate” by smkrv. Read the original source.

This expanded guide is an AI-narrated adaptation of the source’s essential explanation, examples and caveats, not a word-for-word reading. The synthetic voice does not imitate the author or imply their endorsement.

License: MIT

MIT License

Copyright (c) 2026 SMKRV

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.

Our summary

This project helps you test Jev, an AI tool that chooses from options rather than writing an answer. People use it to see if the software can reliably sort information, like figuring out which department should handle a customer message, before relying on it for real work.

You provide a set of practice examples with the correct answers already marked. The software asks the AI the same questions multiple times to see if it changes its mind. You can then adjust your instructions based on the mistakes it makes during these practice rounds.

This method is useful for anyone organizing incoming messages. To keep the final test fair, the system records which examples you have already used. The guide warns that the short codes used for this record might accidentally reveal private information if someone guesses the original text.

Key takeaways

Check your instructions against practice examples before paying for the AI to process real tasks.
Ask the AI the same question several times to see if its answers remain consistent.
Keep your final test examples separate so you do not accidentally design instructions just to pass them.

More guides →See the experiments →