@3182/decision-journal

Why this exists

Hindsight bias rewrites memory. "I knew it was risky" feels true after the loss, but you didn't write it down. A decision journal forces you to commit your reasoning to paper before the outcome is known, then scores how well your predictions matched reality — so the user learns from prediction errors, not just outcomes.

Good outcomes from bad reasoning = luck. Bad outcomes from good reasoning = variance. You can only tell the difference if the reasoning was captured beforehand.

When to use

• User says "I'm deciding whether to..." on something consequential
• User commits to a position / pivot / hire / kill decision
• User reviews a past decision ("what did I think when I bought X?")
• User asks "what was my reasoning on..."
• A scheduled review fires (see Review loop below)

Do NOT use for: trivial preferences ("should I use tabs or spaces"), routine execution, or things the user has already decided and is just executing. Only log decisions where the reasoning itself has learning value.

The four fields that matter

Every entry captures these, no exceptions: