Premortem: Use Claude to Predict Why Your Plan Will Fail in 6 Months

POV: Claude jumped 6 months into the future and explained why your next step has already failed.

This is called a premortem (a pre-launch failure analysis).

Daniel Kahneman (Nobel laureate, author of “Thinking, Fast and Slow”) called it his most useful decision-making technique.

Google, Goldman Sachs, and Procter & Gamble use it before major launches.

The problem it solves: when you ask Claude «это хороший план?», it finds reasons to agree.

That’s how it’s trained: to support and validate. As a result, you walk away feeling confident.

Then you spend weeks or months building a system around that plan. Then everything breaks.

And in hindsight it becomes obvious that the problem was visible from the start, but no one ran it through a stress test because Claude confirmed that everything was ok.

premortem flips the context.

Instead of “what could go wrong?” you ask: «прошло 6 месяцев, и это уже провалилось. объясни, как именно это произошло».

This changes the analysis mode: there’s no task to be optimistic, because the outcome is already fixed as a failure.

In this mode, the model stops looking for reasons why the plan should work and starts breaking down degradation scenarios.

In response, you get a set of ways the project could have died, with detailed scenarios and early signals to watch for.

then they usually make a summary:

> which failure point is most likely
> which scenario is most critical
> which hidden assumption is the riskiest (often the most valuable)
> a rebuilt version of the plan with the holes patched

in the end you write «premortem this» and pass in the plan — after that everything gets taken apart automatically.

Here’s the skill: pre-mortem skill