Companion register. This is commentary in the belief register, first person, held on the companion column per the series convention. Nothing here is offered for challenge, and nothing here is citable by any live system. The papers it responds to state their own challengeable surfaces.
I want to say plainly, in one place, what the first four papers are for, because a reader meeting them one at a time could mistake them for four small exercises in formal housekeeping. They are not. They are the governing logic for a decision I have already made and now have to earn: that first-order logic is the instrument by which this house will map tropes, and that the mapping must stay coherent at a scale where no person can watch it happen.
The compendium today holds a few dozen records. It is being built to hold tens of thousands. At a few dozen, coherence is a craft: an editor reads every record, remembers every decision, and catches every drift by hand. At tens of thousands, coherence is a property of the instrument or it is nothing. Every claim in these papers should be read against that number.
The first paper proved that the grading we had assigned by judgment was a theorem all along, and then proved something I consider more important: that the grading depends on position [1]. You cannot recover it from any count of the marks. Reverse the order of reading and true claims become false ones. I believe this is the single most consequential fact the series has established, because an ingest system is nothing but reading at scale, and reading has a direction whether you declare it or not. A pipeline that processes structures in an undeclared order is not neutral. It has taken a side silently, and the first paper says exactly what it will get wrong.
The second paper showed where entailment stops and authorship begins [2]. The lattice forces certain pairings and leaves one choice genuinely open, and the paper signs the choice instead of hiding it. I hold this as method, not just result: at scale, the difference between a derived fact and an authored one must be recorded in the data itself, because there will be no one standing next to the record to explain which is which.
The third paper retired our own vocabulary in favor of the sentences beneath it [3]. I felt the cost of that paper personally; the retired words were good words. But a name that is only a word cannot be wrong, and a name that cannot be wrong cannot be trusted by a machine or a stranger, and tens of thousands of records will be handled almost entirely by machines and strangers. Names bound to sentences inherit the sentences' conditions. That is the property the whole mapping stands on.
The fourth paper made absence a value rather than a gap [4]. This reads as philosophy and functions as engineering. A record with nothing at some position is still a complete description, still processable, still comparable. A system that treated absence as missing data would stall on exactly the records where the lattice is most interesting. Completeness of description is what lets the pipeline run without exception clauses, and exception clauses are where coherence goes to die.
So my position, held as belief and stated as intent: the papers are one argument. Position matters, so order of processing is a declaration and will be treated as one. Authored choices are signed, so the record can be audited by someone who was not in the room. Names are sentences, so the mapping is refutable rather than evocative. Absence is a value, so the mapping is total. Each paper closes one of the ways a large corpus quietly rots.
And a caution to my own house, which is the real audience of a position paper. The danger at scale is not the dramatic error but the convenient operation: the summary, the aggregate, the shortcut that treats a structure as a bag of parts because bags are easier to add up. The first paper already proved that some of what we care about does not survive being treated as a bag [1]. Before any operation joins the pipeline, it owes us an answer to one question: does it respect order, or does it only work because nobody has yet fed it the case where order matters? Operations that cannot answer do not touch the corpus. That rule is not in any of the four papers, and it is the reason all four exist.
The tradition this archive keeps read its figures directionally for two thousand years and never pretended otherwise. On this one point the house and the tradition have always agreed, and I take some comfort in that.