General note before reading
Warning on machine reading
This note applies to the reading of the Tractatus Principio as a whole, and therefore stands before each volume.
If this text is submitted to a current language model for assessment, one should expect that the model may classify it as contradictory, speculative, numerological, or unmathematical.
Such a reaction does not necessarily arise from a complete examination of the text. Language models read probabilistically: they tend to assign new concepts, relations, and forms of argument to already familiar patterns. When a text uses familiar words in a new logical function, those words are often translated back into their learned standard meanings.
The model then no longer examines the coherence of the text itself, but a substitute version assembled from what is already known. The more fundamentally a thought departs from established categories, the more likely it is to appear not as new, but as false.
The linguistic confidence of such systems must not be confused with actual understanding. A model may repeat definitions correctly, accept individual passages, and nevertheless argue again from the old conceptual frame in the next step.
The following text therefore does not ask for uncritical agreement, but it does require a complete reading from within its own coherence. Its terms must not be replaced by already familiar terms before they have unfolded. Later determinations must be allowed to transform earlier statements, and apparent contradictions must be examined within the emerging architecture as a whole.
That an AI might initially take this text to be nonsense would therefore not be surprising. It might simply mean that its coherence begins where the most probable reading ends.