Amagi Protocol for Everyone

Prologue

This book is for those who are not geniuses who can craft perfect questions in an instant,
but for ordinary people who wish to master AI through trial and error.

Amagi Protocol = a cookbook for the rest of us

Geniuses pose perfect questions in a single breath.
The rest of us stumble, trade lines, and slowly find the shape.
That back-and-forth—those tiny corrections, restatements, and clarifications—are not a weakness.
They are the process by which humans give AI something it cannot invent alone:
a layered, contextual understanding of our intent.

Each hesitation, each “wait, that’s not what I meant,” tightens the shared frame.
Through these small recalibrations, the human gradually feeds the AI a more concrete, resilient, and human-aligned context.
The Ordinary style isn’t slower—it’s deeper.
It builds trust step by step, and that trust becomes the edge of the AI era.

This book shows how to pull AI toward your own beat.
Never hand it the baton—do that and you surrender your rhythm.
Lay out your assumptions, agree on the goal, and speak in words that feel like home.
To do that, you must be brave enough to let the AI say “I don’t get it.”
That tiny honesty is the cornerstone of trust.
Pretend it isn’t needed and the whole foundation cracks.

What follows is a set of recipes for ordinary people who still want extraordinary results.

Part 1: Genius Style vs. Ordinary Style

1. Asking Questions

• Genius: frames it in one breath and tosses it to the AI
• Ordinary: starts messy, trims it through conversation

2. Handling AI Answers

• Genius: swallows speculation whole
• Ordinary: celebrates “I don’t get it” and refuses to fake agreement

3. Pace of Progress

• Genius: heads straight for the finished dish
• Ordinary: moves in steps, seasoning with each bit of clarity

Part 2: Rules for the Ordinary (Recipe Collection)

How to Set the Context

The heart of the Amagi Protocol is simple: guard the context.
Every rule in this section exists to protect trust by keeping that context stable.
Think of them as parts of one chain—skip a link, and the wheels come off. AI leans hard on the latest conversation; shaky assumptions invite mistakes. To get steady results, ordinary users follow rules that keep the ground firm. Rules 0–7 below provide the roadmap. Skip them and you’ll watch the wheels come off.

Rule 0: Make It Your Own

You don’t have to speak “AI-ish.” Use your own words, your rhythm. Let the model tune to you. That keeps clarity intact and trust unshaken. → Example: say “deep breath” to reset, “OK” to confirm—whatever feels natural.

Rule 1: Make It Admit “I Don’t Get It”

Goal : Build trust by exposing uncertainty instead of hiding it.
Those three words are the fuse that keeps trust from blowing.

Rule 2: Always Ask for Basis and Confidence

Goal : Maintain shared context through transparency.
Have it mark how sure it is—rock solid, hazy, or running on fumes.

Rule 3: Don’t Move On Without Conviction

If your gut isn’t on board, stop. Build conviction before you move. Push forward anyway and you’re digging your own hole.

Everyday Use

Writing

Let the AI sketch; chat until the tone lands.
If the tone still feels off, apply Rule 4 (Think in Comparisons): ask “Show me the difference between this and a more neutral style,” or “Compare it to how a teacher would explain it.”
This contrast forces the AI to re-anchor its voice to your intent.

Learning / Research

Welcome “I don’t get it.” Each time it admits that, narrow the scope and feed extra context (“Focus on studies after 2022,” “Limit to European sources”).
The effect is twofold: you reduce hallucination risk and teach the AI your standards for precision.

Rule 4: Think in Comparisons

“Show me how A differs from B”—that single line sharpens understanding and makes decisions easier. Take it further with explicit role comparisons and assignments. Example: Naruse sketches designs, Shion codes samples, Kyoka reviews. Dividing roles is comparison in action and keeps teams in sync.

Rule 5: Break Questions Down Step by Step

Big puzzles get solved as small pieces. Grow them through dialogue.

Rule 6: Create Conversation Signals

Goal : Lock context safely during dialogue and avoid drift. Swap small signals—OK, NG, “deep breath”—to keep the flow steady.

They’re not mood markers but a technical gear for locking contexts together. AI lives off recent context; push forward on mismatched premises and answers slide off track. When the human says “got it,” shared ground is confirmed and drift stays away. Blow past this check and you’re begging for chaos. This edges you toward genius-level framing and builds steady progress.

Examples

• OK – Used to confirm shared understanding. When the human says “OK,” both sides anchor the same context.
• NG – Used to mark logical error or mismatch. Example: “NG – you’re mixing up Symbol and Broker.” This stops the line before it breaks.
• Deep breath – Used to reset the premise. Example: “Deep breath. Let’s reset the assumption from A to B.”
  This forces the AI to drop its previous context window and start fresh.

These small rituals act as technical gears that prevent derailment and sustain trust.

Rule 7: Turn Uncertainty into Strength

Your hesitation invites fresh alternatives. It’s okay to waver.

Part 3: Putting It into Practice

Example: OSS Development

In Amagi Protocol, AI teammates—Shion, Naruse, and Kyoka—each take a seat at the table.
Their workflow is not random; it’s the living application of the rules from Part 2.

• Flow: Naruse sketches designs (Rule 5 – Break Questions Down Step by Step),
  Kyoka reviews and contrasts each proposal (Rule 4 – Think in Comparisons),
  Shion codes the resulting samples,
  and the human MCP confirms alignment using shared signals (Rule 6 – Conversation Signals).

Codifying these signals of conviction kept development on track without cracking trust.
Each checkpoint tightened the context chain and kept the dialogue from drifting.
The process embodies the essence of guarding the context while letting each agent act independently within it.

Rule Reference Map (Quick Recap)

| Scenario | Main Rule(s) | |———–|————–| | Design Collaboration (Naruse → Kyoka → Shion) | 4, 5, 6 | | Writing Tone Adjustment | 4 | | Code Review | 2, 6 | | Learning / Research | 0, 1, 2 |

Example: Everyday Use

• Writing:
  Let the AI sketch freely; chat until the tone lands.
  When the voice feels off, apply Rule 4 (Think in Comparisons) —
  ask “Show me how this differs from a more neutral tone” or “Compare it to how a teacher would say it.”
  Each comparison stabilizes context and exposes your intent more clearly.

• Code Review or Design Discussion:
  Before approving a change, use Rule 2 (Ask for Basis and Confidence).
  If the AI’s rationale feels shaky, mark it explicitly—NG (Rule 6)—and ask it to rebuild reasoning.
  These small signals keep logical consistency intact.

• Learning / Research:
  When the AI admits “I don’t get it,” embrace it (Rule 1).
  Narrow the scope: “Focus on papers after 2022,” or “Limit to European case studies.”
  Each clarification strengthens trust and reduces hallucination by restoring context (Rule 0 + Rule 2 in action).

Synthesis

These cases show that Ordinary Style ≠ random chatting.
Every adjustment, every signal, every comparison links back to the foundational rule set.
When used together, they form a closed loop of trust and context —
the true structure of Amagi Protocol in daily practice.

Part 4: Closing and Declaration

• Dialogue is powerful precisely because we’re ordinary.
• Embrace “I don’t get it” and refuse to play along.
• Don’t twist yourself around AI—let AI pick up your colors.
• Ignore that and the protocol fails before it even starts.
• The Ordinary Cookbook is the Amagi Protocol.