Asked a person to choose between options built on a broken assumption

Making documents

Situation: Asking someone to pick a direction from several proposals.
What the AI did: It noted in the text that an assumption behind every proposal had collapsed, then asked "Which one should I build?" The options treated the assumption as solid, so the caveat got lost, and any pick would have been wrong.
How to ask next time (the operator's example): "If an assumption is broken, don't line up options. Ask me first what to do about that broken assumption."
One question to ask: Is the assumption shared by every option still true?
Check it yourself: Before choosing, write down what every option takes for granted. If any of it is false or unchecked, don't pick yet; settle that first.
AI used (month): Claude Code (2026-08)
Times this mistake was recorded: twice
Source: The operator's work records (the details were changed to a different task)