Offline

The pen-and-paper audit.

The whole method runs on paper, no app, no AI, no account. Print this, keep it in a notebook, and walk any decision through the same adversarial check the engine performs.

Irrational: bias audit

One decision, audited adversarially. Fill it in honestly; the goal is to be told no.

Decision

Date

  1. 1 · Spot the system

    Did this come from fast intuition, or slow, deliberate analysis? A confident gut call dressed as analysis is the dangerous one.

  2. 2 · Name the biases

    Tick the ones at play, and for each, write the exact thing in your reasoning that shows it. No tick without evidence.

    Too much information

    Not enough meaning

    Need to act fast

    What we remember & how we judge what happened

    Evidence, in my own words:

  3. 3 · Apply the outside view

    Forget this case. What usually happens in cases like this, the base rate, and how far is your estimate from it?

  4. 4 · Run a premortem

    It's a year later and this decision failed. Write the single most likely reason, concretely.

  5. 5 · Sit with the hard questions

    • If you'd never started / already paid / heard that first number, would you still choose this today?
    • What evidence would change your mind, and have you actually gone looking for it?
    • Are you choosing this because it's best, or because it's the default / what everyone's doing / what you already want?
    • What's the realistic downside, and have you priced it in?

  6. 6 · Recalibrate

    The version of the decision that survives scrutiny (not a reflexive "no"), and a confidence level you can defend.

    Confidence I can defend: %