JevMade field notes / Written guide

Design decisions before choosing a model

Augustus teaches a planning method for jobs that might use rules, a classifier, a decision model, or a person. It starts with the outcome that matters and asks whether another judgment would actually change the action, rather than assuming every uncertain step needs AI.

Original by 24601Evaluation

Bookmark Keep this in your collection.

Leave a noteWhat would you try with this?

Only you can see your notes.

Listen to this guide

JevMade’s plain-English explanation

0:00 /

AI narration

Independent ElevenLabs affiliate. JevMade may earn a commission.

Credits and license

“Augustus” by 24601. 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 Augustus contributors

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

Augustus begins by naming the outcome, action, and cost of a mistake. It then asks where another judgment would change the result and whether rules, search, a model, or a person belongs at that point. This keeps the method focused on placement rather than model choice.

The method compares a baseline—a simpler approach used for comparison—and lets the system abstain, meaning decline to decide and ask a person. It builds tests from real labelled cases, measures outcomes, and sets a stopping rule before further improvement work.

This is a design method, not proof that Jev or another model works for a particular job. Teams need representative examples, honest outcome tracking, fixed limits for high-impact actions, and a reason to stop when added complexity does not improve the result.

Key takeaways

Start with the action and its consequences, not the model.
Compare AI judgment with rules, search, and human review.
Decide when to decline, how to test, and when to stop improving the design.

More guides →See the experiments →