Pressure-Test a Decision

• Category: thinking

Description

Get the strongest case against something you've already half-decided, before you commit to it. Counters the default tendency of a model to agree with the framing it's handed.

Prompt

I'm planning to {decision}. My reasoning:

{reasoning}

Don't tell me whether it's a good idea yet.

First:
1. The three strongest objections someone who disagrees would raise. Steelman them — no straw men, no objections you can immediately dismiss.
2. For each, what would have to be true for that objection to be decisive?
3. The single piece of information that would most change your assessment, and how I could get it.

Then give your actual view, and flag which parts of it you're least confident about.

Variables

{decision} — what you're about to do, in one sentence. {reasoning} — why you think it's right. Be honest here; a sanitized version gets you a sanitized critique.

Notes

Withholding the verdict until after the objections is the load-bearing part. Ask for a view first and everything after it becomes justification for that view.

Step 3 is what makes it actionable rather than just uncomfortable — it converts "this might be wrong" into something you can go check.

Works for architecture choices, hiring, and pricing alike. Not much use for decisions that are already made.