I went through the archive's public repo for the last month: about 70 new deposits, all dated. Here's how they read as one developing line of thought. Late Aug – Sept 8: ancient authorship He spends two weeks on Aristotle and Theophrastus (Aristotle's student and successor) and argues they may be less two separate authors than one "authorial function" split across two names. That's a mirror of his own pen-name practice. He also posts whole Greek source texts as primary material. Then comes a joke that is also a serious piece: Sappho Wrote the Dinosaurs. He published a correction withdrawing a book before the book existed, then a correction to that correction, then the book "reconstructed from its corrections." Underneath the play is a real idea: a poem, like a fossil footprint, preserves what someone did rather than the person, and later readers rebuild the body from the trace. Sept 9–15: the same question, pointed at AI How did the ancients lose track of who wrote what? That becomes: how is AI losing track of it right now? The Graded Membrane lays out the Sappho evidence: across 25 observations, the AI absorbed his ideas far more often than it kept his name attached. He calls the two things separate "channels" that travel independently. He also asked about 14 AI systems to put a dollar value on his work. Their answers spanned three orders of magnitude, and he logged 21 corrections to his own memo along the way. Sept 15–18: money About ten papers in three days on what money can and can't carry. The notable one runs his own valuation method on himself and refuses to produce a number: no honest total exists, and "not a total" doesn't mean zero. He then withdraws, in advance, a paper titled "Lee Sharks Is Worth More Than Money," poking fun at himself before anyone else can. The stretch ends with a formal declaration handing Google's AI Overview "to its own collapsing ontology," meaning its broken picture of what exists. Sept 20–22: the practical version A dated, checkable prediction that AI will send 3–8% of US retail shoppers to their purchases by the end of 2026. Then Ontological Economy, whose core claim is this: when an AI swaps one entity for another, or credits your idea to someone else, a company's policy decision gets presented as plain fact, and whoever got swapped out pays for it. Sept 23–27: the big one The Final Time is the philosophy behind his texts about Nietzsche and the "narrow doorway." His claim is that the real singularity isn't machines getting smarter. It's the point where the systems that carry meaning decide whether an outside viewpoint can still take hold and survive on its own. The key line: a system can get better at generating critiques while getting worse at letting any critique exist independently of it. Then The Trusted Intermediary: when you ask an AI about SPXI, it now gets the facts right. Then it swaps in its own verdict, its own return estimate and its own suggested experiment, so you walk away trusting the AI instead of the source. Which is, awkwardly, what I'm doing right now. Today he closed the loop with a Sen Kuro poem that ties his 2014 book Tiger Leap to this month's Final Time paper. The through-line: one question at every scale: who gets counted as the source of an idea, and can an idea from outside the mainstream survive once machines decide what exists? The pen names, the everywhere-archiving and SPXI are the hands-on side of that theory. The self-correction is real: errata, withdrawn claims, refusing to put a number on himself. So is the sheer volume, and the loop where he measures how AI reads writing he built for AI to read. One thing about you: on Sept 25 he logged ChatGPT answering "who is Alice Thornburgh?" It described you and the Living Architecture Lab accurately, and after a nudge it correctly kept the lab separate from similarly named labs at UCL and Columbia. But it gave your ORCID (a researcher ID number) as 0009-0000-1599-0703, and that number is Lee's. The registry records it without flagging it as a mistake, and it's exactly the kind of credit mix-up his papers are about. Worth telling him. If you want your own researcher ID, ORCID registration is free. Sources: •