A decision journal in the Annie Duke style.

Disentangle the decision from the outcome

Most decisions are evaluated by outcome — but outcomes are noisy, especially in the short term. A good decision can lead to a bad outcome and vice versa. The only way to actually get better is to log what you decided, what you expected, and *why*, then revisit when the outcome is known. Annie Duke (Thinking in Bets), Shane Parrish (Farnam Street), Daniel Kahneman — they all recommend a decision journal. The format is well-known. The discipline isn't. Decisions is the calm tool that makes the discipline survive past the first week.

What you see

The **dashboard** shows decisions overdue for review, decisions due this week, your recently logged open decisions, and your recently completed reviews. Every morning becomes "what should I revisit today?" The **decisions** page lists everything — searchable by title, body, expected outcome, or notes; filterable by tag and open/reviewed status. Sortable so the highest-priority review surfaces first. A **decision detail** page holds the original capture (what + why + expected outcome + confidence) and the full review history. Logging a review is structured: what happened, what you learned, and a 1–10 hindsight score *separate* from the outcome. The **stats** page shows your hindsight-score histogram, your **calibration curve** (high-confidence decisions should hindsight-score higher than low-confidence ones — if not, you have a calibration leak), and a by-tag breakdown so you can spot domains where your judgment is sharp vs blurry.

What it doesn't do

No AI summaries. No "decision templates" to fill out. No team workflows. No public sharing. No outcome scraping. Decisions is single-user manual entry — the act of writing the decision is the act of making it.

Built to compound over years

A 1–10 confidence score *and* a free-text confidence word ("leaning yes", "coin flip"). A 1–10 hindsight score *separate* from outcome. Comma-separated tags for lightweight grouping. Markdown-friendly body, expected outcome, and retrospective fields. Calibration chart that gets more useful every year. Archive without deletion.