Decision Entries and Outcomes

Each entry in Decision Journal holds a decision, the reasoning behind it, and a prediction of how it'll turn out. Once the user adds an outcome later, a Beam agent reads the pair, matches it against the account's rule for what counts as a hit or a miss, for example a confidence score that missed by a wide margin, and updates a running record of accuracy. Entries still waiting on an outcome are left alone. Ones where the outcome is ambiguous, or where the original prediction wasn't recorded clearly enough to score, get left for the person to judge instead of forcing a verdict.

Prediction Accuracy Patterns

Over enough entries, Decision Journal shows a picture of where someone's predictions tend to lean optimistic or pessimistic. A Beam agent reads that pattern against the account's rule, for instance flagging a category of decision where the gap between prediction and outcome keeps repeating, and notifies the user with a short summary rather than a raw list of entries. Genuinely new patterns, or a shift severe enough that the rule doesn't clearly cover it, get left for the person to interpret. The aim is to surface a repeating blind spot early, not to grade every single decision the same way.

Decision Tagging and Categories

Entries can be tagged by category, like hiring, spending or scheduling, so patterns show up within a type of decision rather than across everything at once. A Beam agent reads new entries as they're tagged, applies the account's rule for grouping or summarising a category, for example a monthly note on how spending decisions have gone, and updates the summary the user sees. Entries tagged inconsistently, or that could fit more than one category, get left for a person to sort out. This keeps the categories meaningful instead of turning into a catch-all tag nobody checks.