PAP0001activev1Categorical AI
Presents a substrate-aligned framework for reading modern AI architectures through the τ-Kernel construction. Includes falsifiable predictions about hallucination, in-context generalization, and architectural reliability.
Payload
Canonical artifact:https://panta-rhei.site/publications/papers/categorical-ai/pdf
Abstract
This paper presents a substrate-aligned framework for reading modern AI architectures (transformer, mixture-of-experts, recurrent, hybrid) through the τ-Kernel construction. We articulate prediction families around hallucination as patch failure, in-context generalization as finite-budget coherence extension, and architectural reliability as a function of substrate-alignment depth.
Citation
Fuchs, Thorsten, and Anna-Sophie Fuchs. Categorical AI: A Substrate-Aligned Framework and Falsifiable Predictions. Panta Rhei Research Program, 2026.
Identifiers
Aliases & legacy IDs
categorical-aiExternal identifiers
Release lines
corpus_v3_workingVersion & History
Status disclaimer
A Corpus Item page reports the program's current internal record for this item. It does not imply external verification, scientific consensus, or final proof unless explicitly stated. Read it together with its dependencies, formalization status, and the program's overall stance.