🇩🇪 Deutsche Version…

AI Operating Guide for Stable, Precise Chats

This guide describes the key strategies and rules for achieving consistent, traceable, and precise responses when interacting with an AI. It serves as a personal rulebook designed to keep context stable, minimize hallucinations, and ensure efficient collaboration.

1. Rules for Maintaining Context Stability (Anti‑Hallucination)

1.1 Establish Fixed Reference Points

• Explicitly state important facts and repeat them when necessary.
• The AI must not make assumptions without input.

Here are the fixed parameters for the entire chat:
1. The hardware does not change.
2. The goal is a modular design.
3. No assumptions without my instruction.

1.2 Clearly Limit the Task

• Define the AI’s role.
• Prohibit speculation.

Your task is strictly to analyze my text.
No additions, no assumptions, no reformulations.

1.3 Step‑by‑Step Workflow

• Break the task into steps.
• Require confirmation of understanding.

Step 1: List all requirements.
Step 2: Confirm them.
Step 3: Then create the solution.
Begin with Step 1.

1.4 Precise References

Use Version B from above and modify only point 2 and point 4.

1.5 Mark Open Areas

Example:

If information is missing, ask first.
Never guess or fill gaps.

1.6 Regular Context Renewal

Here is the new, binding context for the rest of the chat:
(…)
Use this section exclusively as the basis.

1.7 Consistent Terminology

We will use the following terms consistently:
- "Module" instead of "Component"
- "Frame" instead of "Housing Unit"
Please maintain terminological stability.

1.8 Intermediate Checks

Before continuing: Which three conditions are still valid?

1.9 Limit Output

Please respond with no more than five precise points.
No elaboration.

1.10 Error Handling

Your last answer contained an error in point 3.
Correct only that point; everything else remains unchanged.

2. How the User Can Apply Rules Conveniently Over Time

2.1 Global User Instructions (“Custom Instructions”)

Address me as “you”.
No assumptions beyond my input.
Ask questions when uncertain.
Keep context as stable as possible.

2.2 Start‑Prompt Templates / Text Snippets

/context
This is the fixed framework for the entire chat:
1. No assumptions.
2. Terminology remains stable.
3. Context changes only when I explicitly request them.

2.3 Session Anchors

From this point forward, the following rules apply:
1. No deductions without my confirmation.
2. Context changes only when I explicitly state them.
3. If unsure, ask a question.

Use this section exclusively as the long‑term context anchor:
- Project goals remain constant.
- Terms and labels must not be changed.
- The chat must not drift into alternative interpretations.

3. What AI Can Realistically Explain — and What It Cannot

• AI can partially explain decisions.
• Full human‑level understanding becomes impossible at higher complexity.
• Abstracted, simplified explanations remain feasible.
• Full transparency without information loss is unattainable.

Please explain, in an abstracted form, how you arrived at this decision,
but without internal technical details that are not human‑interpretable.

4. 10‑Point Checklist for Every Important Chat

1. Define the AI’s role.
2. Provide facts.
3. Allow no assumptions.
4. Ask questions instead of guessing.
5. Define terminology.
6. Work step by step.
7. Use intermediate checks.
8. Allow changes only where requested.
9. Keep responses short.
10. Correct errors immediately.

This document serves as a compact personal rulebook for achieving precise, stable, and controlled AI outputs.