Person/Nathaniel Whittemore/Pod/26/11 How to Deslopify Your AI
Properties
- date-created
- 2026/11/02
- Person/Nathaniel Whittemore/Pod, with guest Person/Nufar Gaspar. Gaspar describes how to brief an AI tool, ask it to challenge your thinking, and check that its draft fits the intended reader.
Before you open the tool
- 27:22 — Think Before You Open The AI Tool
- Gaspar's first habit: "you have to know what you're thinking before you open the tool." Take a walk or read until you have a view of your own, however messy.
- Then brain-dump to the tool by dictation: what you heard, what worries you, what you think, and what you are sure you don't want. Let the tool find the structure. She says this episode started with 15 minutes of her talking to an AI tool that way.
- She calls this the step people skip that has the largest effect on everything after it.
- 28:13 — Tell AI How Hard To Challenge Your Thinking
- Tell the model how confident you are and how hard to push back, because it is built to agree with you. Her example: "I'm about 60% sure the problem is onboarding. Push back hard before you agree with me."
- Before starting, decide who the audience is and what you want them to feel or do, with standards for length, format and tone. That target also gives you something to check the result against.
While the tool works
Review
- 31:16 — Convene An AI Review Board
- Before important work goes out, have AI review it as distinct readers: a skeptic, a domain expert, the busiest recipient, and a devil's advocate who argues it should not go out at all. Use only the feedback you find valuable.
- 31:55 — Respect Your Audience's Read Budget
- Never send something you wouldn't read yourself. Otherwise the reader hands it to their own AI to summarize, "and then now we have two models that are talking while two people pretend to communicate."
- Distilling a page into five sentences forces you to decide what you think. Choose the format on purpose.
Sources