Against Intelligence-First

Prime 3 clears intelligence. It corrects a recurring interpretive failure: treating "intelligence" as a primitive essence, a synonym for complexity or optimization, or a private, human-centric possession — smuggling in consciousness, personhood, or magic wherever the word appears. The paper de-privatizes the term by defining it operationally: a system is intelligent when it can update its own constraints, using plastic memory and feedback, such that it improves prediction or compression, control, and transfer across contexts. On this view, intelligence is not more coherence and not more complexity — it is coherence that becomes learnable.

The structural criterion is an endogenous update loop that modifies constraints in response to evaluative feedback and yields measurable gains that persist beyond the training context. "Endogenous" qualifies the mechanism and memory, not the signal: feedback may come from the environment, but the plastic state change must be owned by the system. This turns intelligence claims into something falsifiable. The paper supplies a minimal litmus test, a feedback ladder that separates mere regulation from genuine learning — the thermostat does not qualify — and a compact reporting metric that makes claims comparable within declared baselines and evaluation suites. The discipline is to state the non-learning baseline you are beating, and the transfer suite on which you beat it. As with every Prime, the paper asks no commitment to Universal Collapse Theory — intelligence is a term the reader already holds and can judge on its own ground.

What this paper does not claim. The paper does not claim that an intelligent system is conscious, that it understands its task, or that it has inner experience, personhood, or rich agency. Nor does it claim that complexity, large state spaces, or optimization-shaped behavior are themselves evidence of intelligence, or that intelligence is uniquely human. It authorizes no law-level claims; it is a reader lens, held as provisional and open to revision.

Keywords: intelligence; plasticity; feedback; transfer; conceptual hygiene.