Collation and the microscope: how agentic stemmatics and interpretability could check each other

Interpretability asks why a model does what it does. Agentic stemmatics asks which models share errors or ancestry, and how much their agreement is worth as evidence. This research note argues the two could check or sharpen each other: internal features offer a second measurement of the descent signals the collation method relies on, and the philological reasoning about shared errors, completed with the probability under descent, the term that likelihood methods in phylogenetics carry, suggests how to weigh similarities between models as evidence of relationship rather than of convergence or other common causes. An experiment was drafted and put through six rounds of review by models from two families other than the one used in drafting (Codex, OpenAI; Kimi, Moonshot AI); the reviews did not converge on a design. The note publishes the eight conditions a valid test of descent from shared characters has to meet, as currently stated, and moves the design to a pre-registration. This is a proposal, not a result. Files: the PDF (a print of the HTML page), the HTML page, the Markdown reading copy, and the review-records supplement (zip), which holds the six round records, the brief and packet each reviewer received, the seat outputs, the same-lineage screening passes and all sixteen drafts. Version 1.0 corresponds to the site copy of 16 October 2026 at research.counterproof.io/note/collation-and-the-microscope.html. Companion to Agentic Stemmatics (doi:10.5281/zenodo.22030516; the text cited is v1.38, doi:10.5281/zenodo.23424032). Drafted with a Claude model (Anthropic) and revised by the author; same-lineage screening passes are recorded in the supplement and not counted as independent review.