Paper outlines a five-level communication architecture for language-based ASI

The proposal separates systemic activity representations, situation schemas, narrative, full sense and text, arguing context alone cannot expose utterance foundations.

The paper argues that a language model's context cannot present the foundations of utterance formation for examination, because those foundations come from acquired schemas and systemic representations of activity. It defines five levels: systemic representations, a situation cut from them, a narrative description, the full sense of a message, and text. A recipient is said to reconstruct full sense from abbreviated text through jointly agreed representations. Reverse work runs from full sense back to systemic foundations. The account draws on Shchedrovitskiy, Korzybski and Mel'chuk. Further research is described as large graphical models of activity rather than context-based text generation.