JevMade hello@JevMade.com
← Back to guides

JevMade field notes / Architecture guide

Map out future choices to help an AI assistant plan ahead

This guide explains how to use software that maps out possible choices for an AI assistant. It helps the software look ahead and pick a complete path instead of just guessing the immediate next move.

Original by Chuf-HAgent workflows

Listen to this guide

JevMade’s plain-English explanation

0:00 /

AI narration

Credits

“Build an executable probability tree from local Jev decisions” by Chuf-H. Read the original source.

This expanded guide is an AI-narrated adaptation prepared by JevMade. It expands the source’s essential ideas, examples and caveats in JevMade’s own words and is not a word-for-word reading. The synthetic voice does not imitate the author or imply their endorsement.

Our summary

JevTree is software that helps an AI assistant plan ahead. It uses Jev, an AI tool that chooses from options rather than writing an answer. People use this to solve complex problems where the most obvious immediate choice might lead to a dead end later.

The software asks the AI to score different allowed choices for the current step. Instead of taking the highest score right away, it groups these questions together to map out many future paths. It multiplies these predicted scores to find the path most likely to succeed.

This method is useful for developers building tools with clear rules and predictable steps. The AI only makes choices, so normal code must still enforce the rules and check if the task is finished. Multiplying predicted scores is just a planning method, not proof of real-world success.

Key takeaways

  1. Write clear rules in your code so the AI assistant only explores allowed and possible actions.
  2. Group questions together to map out future paths instead of just picking the immediate best choice.
  3. Use regular code to check if the final goal is met after the AI makes its choices.

In these tests, mapping out paths worked better than picking the immediate best choice, but this is not a general ranking of AI models. The software requires an API key, which is a private access code, and relies on regular code to enforce rules.

GitHub README and planner implementation · Source reviewed

Read the original guide Opens the author’s site in a new tab.