3.1The primary text
The text is Max Weber, Politik als Beruf, in the first edition (Duncker & Humblot, München und Leipzig 1919), as transcribed and twice proofread by German Wikisource and fetched on 2026-10-05, with the printed page numbers removed. The copy used gives three passages: the three inner justifications of rule, with the lecture's interest in rule by charisma; the passage setting the official beside the political leader; and the long final section, from the two maxims of ethically oriented action to the end of the lecture. All quotations of Weber are copied from that copy, in German, character for character, including its quotation marks and its spelling; where a page break in the edition divides a sentence, the quotation stops at the break. The transcription prints "veranwortungsethisch" in one place where the word is elsewhere spelled with a t; the quotation keeps it as printed. Every English rendering of Weber in this article is the article's own translation, and is marked as such where it first appears in each section.
The lecture is quoted at length because the argument depends on its wording. Weber's passages on responsibility use a small number of verbs with precision, and a paraphrase that smooths them would erase the distinction the article follows.
3.2The record
ARCHON publishes small scenes drawn from working records, anonymised and summarised, each read through eight lenses. Each lens poses a question, sets out three options with what each gains and risks, and gives a judgment marked as an interpretation open to revision, dated, with a condition for reconsidering it. The raw records are never published. The scenes are synthetic in the sense that matters for this article: they are modern summaries written for a platform, with roles in place of people, and every reading attached to them is the platform's. They are not archival documents and cannot be used as evidence of what any particular organisation did or what any person thought.
The scene read here is titled, in English, "Order form received automatically but not split", and is called here Unsplit Order Form. Its setting is dated 2023; its readings are dated 2026-09-20. It is published in Korean, which is the original, and read in an English translation that carries the hash of the Korean file. The first lens is marked as inspired by the record; the other seven are marked hypothetical, which means they suppose a condition and never report what happened. Across the ten scenes the study reads closely, the judgments of the seven hypothetical lenses follow one template, with the roles and the missing element changed. The repetition is a property of the platform's writing and is not treated here as a finding about the scene. Where the article quotes the record, it gives the lens and the field.
3.3The ledger
The study keeps a decision ledger over 186 of ARCHON's cases. It was made in two steps. Code first collected what the records state directly, such as whether a condition for reconsidering the judgment is present and whether the record names a decider. An automated reader was then run twice, independently, over each record, answering a fixed list of questions. One list concerns the study's general axes: who decides, who bears the cost of waiting, how far the decision reaches in time, how hard it is to undo. A second list, registered in advance on 2026-10-05, concerns restraint: whether the record shows one, where it comes from, whether the scene's own words give its reason, whether a condition for removing it is named, whether those who wait under it could invoke that condition, and whether they can see what it protects and how far it reaches. A cell counts as settled only where both runs agree. A strict reading counts only what both runs found stated in the record's words; a lenient reading also counts what the setting implies.
The restraint questions were first coded with the platform's own list of what it cannot establish included in the reader's material. A further check showed that the reader took that list for a restraint, and the questions were coded again without it. Only the second coding is used in this article. On the question this article most depends on, who bears the cost of waiting, the two runs agreed on 175 of 186 cases (94%). On who decides, they agreed on 131 of 186 (70%), and the records themselves name a decider in none of the 186. These figures are reported in plain words in Section 6 and in the study's terms in the Appendix.
3.4How the cases are used
The two readings of Weber and the one reading of the record are cases of two kinds George and Bennett set out: heuristic case studies, which use cases to find new variables, paths and questions, and plausibility probes, preliminary studies that ask whether an untested idea is worth fuller testing (George and Bennett 2005). Here the idea is a distinction found in a text of 1919. A lecture and a synthetic scene can show that a distinction exists and where it bites. They cannot show how often it occurs in real organisations.