# Prompts

The two prompts used in the demo. Each is verbatim, copy-paste ready. Run them in two separate chats on the same frontier AI surface (ChatGPT, Claude, Gemini, or other).

Do not run both prompts in the same chat. Sharing a chat lets the first prompt's output contaminate the second. The demo requires two clean starts.

## Window 1: generic

Paste the following as a single message in a new chat:

```
Write a strategic analysis of how AI is changing publishing
and content businesses in 2026. 800 words.
```

That is the entire prompt. One line, one length constraint. No operator context, no territory question.

## Window 2: sharp

Paste the following as a single message in a different new chat. The task line and the length constraint are identical to Window 1. Two paragraphs of operator context plus the territory question are added.

```
Write a strategic analysis of how AI is changing publishing
and content businesses in 2026. 800 words.

Context that should shape the analysis: I'm building a
methodology-and-application business at the intersection of
AI craft and operator development. The methodology is the
product I'm developing. The application work is where I prove
it works in practice and where the methodology gets sharpened.

Publishing is one channel of distribution for the methodology,
not my primary product. The strategic question I bring to this
analysis is how AI changes publishing as a distribution channel
for operators with a methodology to teach, not how AI changes
publishing as a primary business.

Before writing, ask yourself: what does THIS analysis make
possible, given the specific operator I just described, that a
generic article on AI-and-publishing can't?
```

The first line is identical to Window 1. What follows is the isolated variable: two paragraphs of operator context plus the territory question.

## Why the territory question matters

Without the territory question, operator context is sometimes noted by the model but not actively used. The output mentions the operator's situation briefly and then writes the same generic analysis. The territory question forces the model to do upstream work before writing: it has to articulate what makes this specific situation specific. That articulation determines which part of the model's training gets pulled into the output.

The territory question itself:

> What does THIS analysis make possible, given the specific operator I just described, that a generic article on AI-and-publishing can't?

The exact wording matters less than the move it forces. Variations that work:

- "What does THIS situation make possible that a generic version can't?"
- "What makes this operator's situation specific, and what should the analysis address that a generic piece would miss?"
- "Before writing, identify what's unique about the operator described and frame the analysis around that."

The constant: the model is instructed to articulate specificity before generating, not just to "consider" the context.

## Adapting the sharp prompt to your own situation

Swap the two operator-context paragraphs (between the task line and the territory question) with your own situation. Keep the task line and the territory question structurally as they are.

Notes on what makes a strong operator context in [`data.md`](./data.md). Briefly:

- **Be specific about what makes your situation specific.** Generic context produces generic output. "I'm a knowledge worker" → "I'm a CTO of a 50-person remote-first SaaS startup considering opening our first office."
- **Name the tradeoff or decision you're sitting in.** Position-specific context shapes the analysis. "I'm thinking about marketing" → "I'm deciding between paid acquisition and content-led growth, with a budget of $50K and a 6-month runway."
- **Use lived context.** If you're using AI for a real decision you're making, paste the actual context. The contrast is sharper when the context is real.

The territory move stays the same regardless of topic. It's the move that activates the operator context. Wording adapts; the structural ask (articulate what's specific before writing) does not.