Red-Team My Judgment

Find only what is most likely wrong — no upsides

Here is my judgment: [my conclusion]. Do one thing only: find what is most likely wrong with it. Requirements: 1) Give at least one counterexample or contradicting data point; 2) Name the variables I may have missed; 3) Say when the error would first surface if the judgment is wrong. Do not reassure me, and do not add the merits.

How to use

Keep your conclusion to one sentence; give it three paragraphs of reasoning and it will critique the wording instead of the claim.

Good for

The conclusion has already turned into action — a contract is in legal, a role is being hired, the order is placed — and only now are you doubting it.
The night before a pitch, you've read your own deck three times and it still looks airtight, and nobody on the team has raised an objection.

Not for

The idea is still a vague direction and you can't write a single sentence that could be proven wrong — it will critique your wording instead of your claim, and all you get is copy edits.
You've already decided to do it and only need help filling in execution details and risks — use a pre-mortem instead; red-teaming at that point just makes you oscillate at the starting line.

Example

Input

I think cutting the free tier from 100 requests a month to 30 will double our paid conversion. Basis: a competitor did the same last year and their paid rate went from 2% to 3.8%.

What you get

What you want is a specific counterexample, not “think it over”: the competitor also redesigned its pricing page and added sales outreach that quarter, so the lift may not have come from the cap at all; the variable you missed is that 60% of your free users only ever run one request, so the cap is invisible to them. It should also say when the error surfaces first — new paid conversions in week three, by which point free signups are already falling.

Common mistakes

Using it to seek agreement rather than find errors: when an objection doesn't suit you, you rephrase the question and ask again until the model says the judgment is fine.
Pasting only the conclusion with none of the facts behind it — it can't find the real weak point, so it critiques the shape of the logic or invents a counterexample for you.

Try tweaking it like this

Swap “find what is most likely wrong” for “give the strongest opposing argument” — you move from poking single holes to forcing a full alternative explanation.
Add “rank these by how much it would cost me if this one is wrong” at the end — attention shifts from where it might be wrong to which error would hurt most.