Building a temporal layer so your AI never lies about time

Ahnii! Series context: This post builds on the Waaseyaa series. Claudriel is an AI personal operations system built on the Waaseyaa framework. You don't need to have read the earlier posts, but they cover the entity system and architecture that this temporal layer sits on top of. Most applications treat time as a free function call. Need the current time? new DateTime(). Need it again three lines later? new DateTime() again. In a request that takes 200ms, nobody notices the two-millisecond difference between those calls. An AI system that reasons about your schedule, detects drifting commitments, and nudges you before meetings does notice. If the commitment extractor captures "now" at 14:00:00.003 and the drift detector captures it at 14:00:00.217, you get inconsistent temporal reasoning. Worse, if the system clock drifts from reality and nobody checks, every time-based decision is quietly wrong. This post covers Claudriel's Temporal subsystem: how it pins time per request, resolves th