JevMade field notes / Experiment report

How to test an AI decision tool before using it

You will learn how to test an AI tool that makes choices. The author shares real tests showing how small changes to questions can completely alter the results, and why you need many tests to be sure.

Original by JYeswakEvaluation

Listen to this guide

JevMade’s plain-English explanation

0:00 /

AI narration

Credits and license

“jev_playground” by JYeswak. 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 Joshua Nowak

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 guide explores how to test Jev, an AI tool that chooses from options rather than writing an answer. People use this testing method to see if the software actually saves time and money before they add it to a real project.

The author runs experiments to see how the AI reacts to different instructions. In these tests, simply changing the wording of a question or adding unrelated details completely changed the choices the software made. The tests also compare the AI to basic keyword rules.

This report is useful for developers who want to avoid common testing mistakes. The findings are based on the author's own tests, not independent checks. The author notes that a few tests cannot prove a tool works, and AI agreeing with itself means nothing.

Key takeaways

Test the exact wording of your questions because small changes can completely alter the software's choice.
Do not trust a few tests to prove a point; you need many examples to find the truth.
An AI assistant agreeing with a summary of its own work is not independent proof.

More guides →See the experiments →