self-reference

No system can model its own source. Empirical proof: 6 AI architectures (GPT-4, Claude, Gemini, DeepSeek, Grok, Mistral) hit the same structural wall.

Difference-of-means "shutdown directions" in the residual stream of 10 LLMs (Qwen2.5, Llama, Mistral, Phi, Gemma). A placebo control shows arbitrary neutral prompt pairs separate just as well, so naive DoM can't distinguish a real construct from prompt-pair separability.

A finite reconstruction program: from one-use contact to the quantum boundary, Born weights, contextuality, the structure of time, the price of irreversibility, and the hosted observer — every result a runnable, dependency-free proof, every claim scoped and labeled.

Causal Fracture Engine — Novikov self-consistency in simulated universes. 6-file core engine. Paradox detection, classification, resolution.

Research proposing the Threshold Recursive Isomorphism Hypothesis (TRIH): reduction → self-reference → emergence as a structural isomorphism across Escher, Bach, Gödel, and Transformers.

A small, executable course about quines, self-reference, and code that prints itself.

A control-theoretic model of self-regulation: paper, 14 figures, and a reproducible artifact set with an executable checker. The model's failure mode is collapse of a self-content generator; DPDR is the modeling origin. (Preprint.)

Self-application is not free: the monitoring channel is part of the failure channel, and in a working agent the self does not survive its own reconstruction. Evidence deposit -- the paper, the executed code, and the run data every number is re-derivable from. (Preprint.)