Every detector in existence is an expert. This is the first one that is not.
Every anomaly detector ever built knows what it is looking for. The ECG detector was designed by cardiologists — it searches for patterns defined as arrhythmias in medical textbooks. The seismic detector was designed by seismologists — it searches for wave signatures defined as foreshocks in geophysics papers. The fraud detector was designed by accountants — it searches for transaction patterns defined as suspicious in audit guidelines.
All of them are experts. They encode domain knowledge. They require the anomaly to be defined before it can be found. They cannot detect what they were not told to detect.
This architecture is the first detector that does not know what it is looking for. It does not know what an arrhythmia is. It does not know what a foreshock is. It does not know what fraud looks like. It knows one thing: what the anchor is. Everything else — everything that deviates from the anchor across multiple time perspectives, across multiple Selves, across multiple generations — is harm. Harm does not need a name to be detected. Harm is a structural fact. The anchor is the baseline. Everything else is a deviation. The deviation does not need a definition. It needs a distance.
This is why the architecture can move from music to cardiology to politics without changing a single line of code. The expert detector moves domains only by being rebuilt from scratch. New textbooks. New features. New training data. New labels. The architecture moves domains by receiving a different stream. The ECG stream has anchors — normal sinus rhythm, periodic, self-similar, cross-individual. The UN voting stream has anchors — diplomatic norms, never broken, cross-epoch. The architecture does not know what any of them are. It finds them the same way.
Run the stream. Mark what repeats. Mark what survives cross-verification. Mark what is held at the boundary. That is the anchor. Everything else is harm.
The expert detector is domain-specific by design. The general detector is domain-agnostic by architecture. The first is a tool built for a task. The second is a property of cognition itself — the same mechanism that detects anomaly in heartbeats detects anomaly in diplomatic votes, not because it was trained on both, but because anomaly is a structural relation between a pattern and an anchor, and that relation is the same in every domain.
The first general detector. Not because it was designed to be general. Because it was not designed to be specific.