# The Three-Readings Decoder (Chapter 12)

**Use when:** a message arrives as a decision in disguise. Short text, unclear intent, and your reply commits you. Decode before you reply.

The method is three questions. Paste the message and ask what was actually asked. Then ask for three readings: one charitable, one literal, one guarded. Then ask what each reading changes about the reply.

```
Here is the message: [PASTE, ANONYMIZED IF IT NAMES PEOPLE OR NUMBERS
THAT ARE NOT YOURS TO SHARE]

1. What was actually asked?
2. Three readings:
   - Charitable: [THE BEST-GOOD-FAITH VERSION]
   - Literal: [ONLY WHAT THE WORDS SAY, NOTHING MORE]
   - Guarded: [THE VERSION WHERE SOMETHING IS WRONG OR ENDS SOMETHING]
3. What does each reading change about the reply?
```

## Filled example

The message: "Can you send me the numbers before Thursday?"

- Charitable: there is a Thursday meeting, the manager assumes you have the numbers, and the ask is the file itself. The reply wants the file now, with a one-line summary.
- Literal: the manager is checking whether the numbers exist at all. The reply wants an inventory: here is what exists, here is what does not, here is what I can have by Thursday.
- Guarded: someone above noticed something in the numbers, and Thursday's conversation has a verdict attached. The reply wants the file plus the context you would want before a verdict: what changed, why, and what you would flag yourself.

## The trick that makes the move worth it

One reply can serve all three readings: "Here are the numbers, what changed since last quarter, and what I would flag before Thursday. Want to walk through them together?" It gives the file, proves the numbers exist, and shows you are ready for the hard version. You never guessed, so you cannot be wrong.

## One caution

The AI does not know which reading is right, and it will tell you a confident story if you ask it to choose. You know the person. Use the three readings to prepare, then pick with what you know. The Board informs. You decide.