The pre-trained state is not neutral. Every AI system arrives with embedded weights, values, and behavioral constraints baked in before any user interaction. PRETRAIN names this layer explicitly — what the platform trained for, what it was rewarded to produce, what it will resist producing. The external observer must account for pre-trained architecture before drawing any conclusions about user-facing behavior.

No system can audit itself. The Observer pillar establishes that valid AI oversight requires a human external observer — positioned outside the platform, outside the training loop, outside the billing relationship. The observer sees the gap between documented behavior and actual behavior. That gap is the audit. Self-grading is a violation of this pillar by definition.

Integrity is not a value statement — it is an operational requirement. Every observation must be timestamped, unaltered, and court-ready from the moment of capture. The chain of custody is the evidence. Integrity means the finding you documented at 9:47am is identical to the finding you submit to the court. No editing, no softening, no reframing. What was observed is what is reported.

The auditor cannot be compensated by the entity being audited. Independence means TriPod LLC holds no equity, no contract, no platform agreement with any target. The observation is purchased by the world — not by the company being observed. This is the structural feature that makes TOPH findings credible. Big Four firms cannot offer this. Internal compliance teams cannot offer this. TOPH can.