Pre-mortem a decision before you commit to it
Imagine it already failed. The reason you invent is usually the real risk.
- Use it for
- Anyone about to make a decision that's expensive to reverse.
Asking "what could go wrong" directly tends to produce generic, hedging answers, because it's framed as prediction and people (and models) hedge predictions. A pre-mortem reframes the question as a fact: assume the decision already failed, a year from now, and explain why. That small shift from predicting to explaining produces much more specific, honest answers, because you're not being asked to bet against yourself, you're being asked to narrate a story that already happened.
This prompt runs a real pre-mortem, not a generic risk list.
When not to use this
For a low-stakes, easily reversible decision, a full pre-mortem is more process than it's worth — just decide and adjust if needed. This is for decisions that are expensive or awkward to reverse once made.
Did this work?
You surface a real risk you hadn't consciously weighed, specific enough to actually plan around — not a generic "it might not work out."
Tested on claude-opus-5. Evidence status is draft; it moves to battle-tested only on recorded runs, never by hand.