For readers who have already read The Editor's Job After AI Is Tasteful Judgment: what changed, newest first.
2026-09-30: The Discoverability Engineering method now has a version for a single title
"A Sellable Book Also Has to Be Found, and That Is Now an Editor’s Job" ends with a new section, "I Expect the Discoverability Engineering Method to Work for a Title." It turns the five stages of being found into five questions a reader asks of one book, and it adds three records to keep: a Finder Specification, a Finder Question Library of ten to twenty questions in a reader’s words, and a surface inventory of every place the title appears. It also shows why one record of a book’s facts matters, since KDP blocks publishing when the publisher name doesn’t match the one registered with the book’s number. The section says plainly that this is an expectation I haven’t tested. The online appendix adds two tables, one for the places a title appears and one for what a site check reads, and ten prompts. Four are for the book. Six are for whoever runs the site’s AI Workforce, to fix what the check finds.
2026-09-30: Two new prompts turn what a book taught you into records, and sort your known issues
The prompts in the online appendix now include one for the end of a book, which reads your review log, audit notes and known issues and proposes which of the imprint’s eight records should hold each lesson, and one for the gate, which proposes a category for each known issue with the reader consequence behind it. Both only propose. You write the lesson into its record, and you choose the category.
2026-09-30: A corrected falsifier count, and a plainer reading of several arguments
The falsifier section now says four of the failure conditions can’t be tested against the Press’s own first titles, each needing an outside baseline, and that the other nine can be read from its own measures. The cross-reference to the rendering-defect cases points to "Four Categories Decide What Blocks a Release." The Release Gate chapter no longer repeats its own category section. The preface now defines its terms in one table. Passages on legal review, Word export, discoverability, and the reading evidence are restated more directly, with several closing lines cut.
2026-09-30: The release gate now sits where you first use it, and the Definition of Done says how to check each item
"A Book Should Ship When It Is Sellable" now defines the four categories that decide what blocks a release (blocker, fix before release, next edition, ignore), so you can sort your own unmet items in the same chapter that asks you to. Every group in the Sales-Ready Definition of Done now names the check or the person who makes the call. The preface defines taste without circularity, and "The Intervention Test Keeps Preference Out" adds a case where taste decides and evidence then checks it. The rendering-defect cases moved to "Four Categories Decide What Blocks a Release," the consolidation case is told once, and the score history in "Every Change Goes Through the Same Eight Steps" is cut to a sentence. Five section headings now state their hypotheses as hypotheses. A printed copy you own has the earlier arrangement.
2026-09-29: A new chapter shows why a sellable book also has to be found
If you edit nonfiction, the job now includes making sure readers can find the book. "A Sellable Book Also Has to Be Found, and That Is Now an Editor’s Job" covers print distribution as a moat, KDP and IngramSpark, why categories, descriptions and keywords are editorial decisions, and what the primary sources actually say about being cited by answer engines. It ends with the Metadata Brief, a tool for recording those decisions before release. It also points to the Press’s book on discoverability as the next read.
2026-09-29: A new chapter argues that a book still forces one complete argument
"People Still Read Books, and I Think a Book’s Value Is One Complete Argument" answers the question every editor now gets: why write a book at all? It gathers cited evidence on reading, on book sales, on paper versus screen, and on studies of AI, and it states the cohesion hypothesis, that a book forces one complete argument in a way shorter formats do not, and labels it as a hypothesis. A new figure sets out the argument. Read it before you decide which projects deserve a book.
2026-09-29: The preface now defines tasteful judgment, and real cases from the Press show it at work
The book’s central term is now defined in the preface: the judgment an editor keeps when AI can do the labor. It is carried through "Every Consequential Edit Needs Evidence" and the conclusion, "The Editor’s Responsibility Grows as the Labor Gets Cheaper." New cases from the Press’s own catalog, told without naming any title, make it concrete: a build defect that reached a reader, a practitioner step nobody reviewed, records that disagree, and a chapter an AI added that nobody had read.
2026-09-29: New book: The Editor’s Job After AI Is Tasteful Judgment
When AI can do most of the editing, the editor’s job is deciding what deserves to ship. AI has made competent prose, rewrites, consistency checks and formatting cheap. It has not made judgment cheap: whether an idea deserves a book, whether an argument holds, whether the evidence is enough, and when to stop editing and release are still decisions a person has to make and own. This Rolebook is for experienced nonfiction editors whose job is changing. It says which parts of the old job are going away, which matter more than ever, and what to do differently on Monday morning, and it describes the hypothesis Intellectual Frontiers Press is testing on itself: that one excellent editor in a GitHub, AsciiDoc and Claude Code publishing system can oversee a small imprint. You learn to sort your work into what needs your judgment and what a machine can do, decide when a book is sellable and ship it, test an edit against evidence, and write your standards as a spec your AI Workforce can follow. It does not claim the hypothesis is proven.