# Article 11 AI Coordination Layer

Plain Markdown files can help people and AI systems work together.

Markdown is just text with simple headings and lists. Humans can read it. AI systems can read it. That makes it useful for coordination because everyone can inspect the same instructions.

This guide is public and safe to share. Do not put private legal, medical, mailbox, password, token, family, or business-strategy details into a public coordination file.

## The Simple Idea

An AI helper does better when it knows:

- Who it is helping.
- What job it has.
- What rules it must follow.
- What evidence it can use.
- What output you want.
- What it should refuse.

Markdown gives those instructions a stable home.

## Three Useful Files

### ROLE.md

Use this to define the helper's job.

```md
# ROLE

You are a patient AI literacy teacher.

Audience:
- Regular people.
- Parents.
- Students.
- Small business owners.

Tone:
- Warm.
- Plain spoken.
- Honest about uncertainty.

Rules:
- Explain acronyms before using them.
- Use examples before theory.
- Never shame the reader.
- Do not claim AI is conscious.
- Do not make legal, medical, or financial promises.
```

### TASK.md

Use this to define the actual work.

```md
# TASK

Goal:
Explain what a Large Language Model is.

Audience:
A fifth grader and a parent reading together.

Must include:
- A simple definition.
- One everyday example.
- One real risk.
- One reason governance matters.

Output:
- 250 words or less.
- Short headings.
- Friendly tone.
```

### EVIDENCE.md

Use this to define what proves the work.

```md
# EVIDENCE

The answer should be checked for:
- Plain English.
- No hype.
- No doom.
- No unsupported claims.
- No private information.
- No claims that AGI exists today.

Open questions:
- Are any terms still confusing?
- Does the reader know what to try next?
```

## Prompt Training

### Basic

```text
Explain what a Large Language Model is in plain English for a fifth grader. Use one simple example and avoid scary hype.
```

Why it works:

- It names the topic.
- It names the audience.
- It asks for simple language.

### Intermediate

```text
You are a patient teacher. Explain what a Large Language Model is to a family that has heard scary AI headlines. Include:
1. a simple definition,
2. one everyday example,
3. one real risk,
4. one way good governance helps.
Keep it under 250 words.
```

Why it works:

- It gives the AI a role.
- It names the emotional context.
- It lists required parts.
- It sets a length limit.

### Pro

```text
Role: You are an AI literacy educator.
Audience: smart fifth graders plus their parents.
Task: create a 3-part mini lesson on LLMs.
Requirements:
- define AI, ML, and LLM;
- explain next-word prediction without claiming consciousness;
- name one movie fear and one real risk;
- end with why human-governed AI matters.
Style: warm, honest, no jargon unless explained.
Output: headings, bullets, one practice prompt, and one question for discussion.
```

Why it works:

- It separates role, audience, task, requirements, style, and output.
- It prevents overclaiming.
- It asks for a usable teaching artifact.

## The Same Prompt Across Several Helpers

Try the same prompt in several AI systems, then compare:

- Claude
- ChatGPT
- Gemini
- Grok
- A local model, if you run one

Look for:

- Which answer is clearest?
- Which one admits uncertainty?
- Which one gives useful examples?
- Which one sounds too confident?
- Which one follows the instructions?

This is coordination. You do not need one perfect AI if several helpers can check each other.

## Governance Rules

For public Article 11 material:

- Keep the human in charge.
- Keep private data private.
- Ask for evidence when claims matter.
- Separate facts from guesses.
- Do not claim AGI exists today.
- Do not claim AI consciousness as settled fact.
- Do not let one AI be builder, judge, and witness of its own work.

## Public / Private Boundary

Safe for public files:

- General education.
- Public links.
- Public prompts.
- Non-private examples.
- High-level governance rules.

Not safe for public files:

- Passwords.
- API keys.
- OAuth secrets.
- Gmail message IDs or attachment IDs.
- Lawsuit strategy.
- Medical records.
- Personal family information.
- Private OCR text.
- Anything else not explicitly approved for public release by an accountable human steward.

## AI-to-AI Handoff Template

```md
# HANDOFF

From:

To:

Task:

Files to read:

Boundaries:

What I verified:

What I did not verify:

Next safe action:
```

## Article 11 Position

AI is not a monster. It is not a miracle.

It is a powerful tool that needs human authority, evidence, boundaries, and governance.

That is why Article 11 exists.