Make Claude Argue the Side You Didn't Pick

A bias-check where, once you're leaning toward a choice, you ask Claude to build the strongest possible case for the option you're rejecting — a genuine steelman, not a token counterpoint — so your decision survives contact with its best objection.

By the time you've decided, you're gathering support, not testing the choice, and an agreeable model will happily reinforce you. Asking it to argue the other side hard breaks that loop: it surfaces the real trade-off you glossed over, the context where your choice is wrong, and the assumption you didn't notice you'd made. If your decision survives the steelman, you can commit with confidence; if it doesn't, you just saved yourself an expensive mistake.

Consequential, hard-to-reverse decisions; times you notice you're already sure; and whenever a choice 'obviously' beats the alternatives — obviousness is exactly when you're most likely blind.

Low-stakes or easily reversible choices where the analysis costs more than a wrong call, and moments when you need to commit and act rather than reopen a settled question.

I'm leaning toward <your choice> over <the alternative>. Argue the opposite as strongly and honestly as you can: make the best real case that <the alternative> is the right call, name the conditions under which it clearly wins, and point out the assumption in my reasoning most likely to be wrong. Don't hedge — steelman it.

Set on building a feature in-house, you ask Claude to argue for buying; it makes the maintenance-cost case you'd underweighted and names the assumption ('our needs are unique') as the weak link — you still build, but you scope it far more narrowly.

Have Claude argue every option in turn as if it were the obvious best choice, each with its strongest advocate. Reading them side by side shows you which case is genuinely strongest versus which one merely matched your prior.

• Accepting a weak, strawman counter-argument and calling it a stress test.
• Using the exercise to endlessly relitigate a decision you actually need to make.
• Only steelmanning after you've already committed resources, when it's too late to matter.