Library / First Principles Framework (FPF) - Core Conceptual Specification
Jump to passage
In this reading

Link to current text

Published source confirmed at last check

Source changed 2026-10-03 10:39:28 UTC · snapshot created 2026-10-03 10:40:04 UTC · last check 2026-10-03 11:40:09 UTC

A.3.3.PI:8 - Common Anti-Patterns and How to Avoid Them

FailureWhy it mattersRepair
Use an aggregate as a predictive state without comparing its merged casesComponents or hidden modes can produce different continuations.Compare those cases and retain their relevant distinction or output range.
Substitute a long-run average for a history-conditioned forecastA decision about the next event can depend on recent observations.Condition on the available history or retain its predictive distribution.
Turn a recovered value into a measurement without carrying its assumptionsAlgebraic reconstruction can amplify errors and depends on the supplied law.Propagate the reading uncertainty and inspect the admitted domain.
Repeat a one-step predictor beyond its supported usePrediction error can accumulate, or the retained output may lack its own update.Construct an adequate state update or a predictor for the required horizon.
Keep a longer history without the intervening inputsSimilar observations can follow different actions and call for different forecasts.Retain the actions that enter the transition account.
Obtain a full hidden-state estimate after a bound has settled the questionThe extra work adds cost without changing the present choice.Use the bound and reopen the distinction when another question needs it.