JevMade field notes / Technical guide

Add focused Jev checks to command-line assistants

jev-axi offers command-line tools for choosing, rating, sorting, and checking text. It reports predictable answers and usage details so an assistant's surrounding software can decide what happens next.

Original by shiftynickAgent workflows

Listen to this guide

JevMade’s plain-English explanation

0:00 /

AI narration

Credits and license

“jev-axi” by shiftynick. 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 Nicholas Underwood (shiftynick)

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

jev-axi is a toolkit of small commands that ask Jev focused questions. They can choose an option, rate material, prioritize computer logs, screen untrusted text, or review code changes. Predictable answers and status numbers let other programs respond without interpreting an AI-written explanation.

Local rules allow routine commands without a model call. Higher-risk text gets separate hazard questions, with action cutoffs kept in reviewable code. Reports show estimates, uncertainty bands, the actual model version, usage, and cost. Reusable answers and spending records are kept locally.

The project's small benchmark did not establish that ranking reduces an assistant's total code-reading cost. Meaning checks can miss hazards or block harmless work, and submitted information leaves the computer. Keep permission checks, direct source reading, and user approval separate from model rankings.

Key takeaways

Use exact local rules before paid judgments.
Keep uncertainty, model identity, and cost visible.
Do not replace permissions or source reading with a score.

More guides →See the experiments →