RCC (Recursive Collapse Constraints) proposes that LLM "hallucinations", reasoning drift, and 8–12 step planning collapse are not training artifacts, but geometric consequences of being an embedded, non-central observer.

Key idea: When a system lacks access to its internal state, cannot observe its container, and has no stable global reference frame, long-range self-consistency becomes mathematically impossible.

In other words: these failure modes are not bugs — they are boundary conditions.

Full explanation + axioms in the link.

For context: RCC is not a proposed fix but a boundary argument. The claim is that hallucination, drift, and short-horizon collapse arise from geometric limits of embedded inference — not from insufficient training or scale.

If someone knows of a theoretical framework that can produce global consistency from partial, local visibility, I would genuinely like to compare it against RCC.

Happy to clarify any part of the axioms or implications.

It woule be really helpful for old retired psychotherapists like myself. for you to be far more congruent when using the term hallucination when referring to LLM's.

An AI hallucination is essentially the model making stuff up. It outputs text that sounds plausible and confident but isn't based on truth or reliable data.

LLM hallucinations are the events in which ML models, particularly large language models (LLMs) like GPT-3 or GPT-4, produce outputs that are coherent and grammatically correct but factually incorrect or nonsensical.

machines do not and cannot hallucinate, thats called: Anthropomorphism.