Kladde 26 FAQ.kladde
What follows is an edited transcript of a conversation between a writer (J) and an AI assistant (C, Claude Opus 5) on 18 September 2026, which produced the Kladde 26 FAQ. It follows on from the conversation that produced the manifesto itself the previous day.
The conversation has been condensed. C’s replies ran a good deal longer than they appear here and have been cut to what changed the document; file transfers and the repeated sending of revised drafts are removed. Exchanges marked as comments were left by J on the draft rather than typed into the chat. Where C checked something on the web, the checking is summarised rather than reproduced.
The first six
J: Identified some FAQs. So far: “I create with AI, and I consider myself the sole author, why do I need this?” — and the answer is that if you feel that describes what you do, that’s absolutely fine. Publishing Kladde workings allows your audience to make up their own mind. “I just don’t want to do this” — nobody is making anyone do it. It’s opt-in. “Isn’t this a lot of effort and overhead?” — it needn’t be, and AI can help. My own Kladde workings have been produced by simply asking the AI agent to produce a transcript of its own chat. “A full transcript is way too long and detailed, nobody wants to read that” — the goal is not to provide a forensic record, but to give the reader enough information to understand how much humanity went into the work. For a novel, it might be sufficient to show the edited transcript for one chapter. “Can’t the workings be easily falsified” — they could, but again, they are not intended as forensic evidence. “Most readers don’t want to read all that” — nor would they be expected to. Just the fact that it’s there may be a reassurance, cf. appendixes or translator’s notes.
Please produce this FAQ as markdown. Add in any other FAQs you think should be there. Ask me the question if you don’t know the answer. Keep it in a strict FAQ format: actual questions, followed by the answer to them.
C: [Writes sixteen questions: J’s six put into question form with their answers kept, and ten more drawn from decisions already settled in the manifesto and the blog post — whether Kladde limits how much AI you may use, whether a proportion must be stated, research versus drafting, whether every piece must carry the mark, reprinting, editing the workings for privacy, responsibility for the model’s errors, the name, whether there is an organisation, and whether this covers anything but writing.]
Two you should look at specifically. I’ve included the responsibility question even though you cut accountability from the blog post, because an FAQ that skips a manifesto clause invites the question anyway. And I answered the privacy one from the fourth clause plus your own precedent — the Crab workings anonymise Malcolm at his request and say so.
Four I couldn’t answer. What is the mark? The manifesto says “not everything we make carries this mark” and never says what it is. Where is the floor — does spellcheck count, predictive text, a grammar tool? Do you have to name the model? And what does a writer under an NDA do?
The mark
J: Perhaps we shouldn’t have used “the mark”, or maybe we keep it and state that “the mark” is averring that you are acting according to, and quoting or linking to, the Kladde 26 manifesto.
C: Keep the concept, but take the word out of the manifesto. A movement needs a noun for the thing you put on a page. The problem is that the manifesto uses it in its second sentence with no referent and never defines it, which is the one thing a spare document can’t afford.
The smallest fix is to name the thing instead of gesturing at it: Not everything we make is published under Kladde 26. Where it is, these are the terms. Then the FAQ defines the mark, which is where someone wanting to adopt the practice will be looking.
The definition wants two parts, because the fifth clause needs both: say so and point to the manifesto, then give the reader a route to the workings. And the self-issuing point is worth stating outright — “mark” carries an echo of hallmarks, which implies an authority, and there isn’t one. Dogme ran a certification body for a while and gave up. Kladde starts where Dogme ended, but only if that’s said.
One consequence: with that definition, the mark and the fifth clause become the same object. A single line reading “Made under Kladde 26 — workings at [url]” is the claim, the citation and the route at once. That’s what a print journal can carry.
J: Yes please, and give the example, just as an example, as used in the footer of my own work: “This work has shared workings, in line with [Kladde](…).”
C: [Adds it.] Your line asserts that workings exist and links to the manifesto, but doesn’t say where the workings are. Fine on your site, where the index puts the link beside the story; a reader who meets the piece anywhere else has the claim without the route. Also, your footer says “Kladde” where the manifesto says “Kladde 26”.
J: Make it say Kladde 26 and put it as the first question.
The catch-all
J: I think the answer to all of those is that Kladde does not specify those things. Since it is not a regulatory body, you are free to interpret it as you see fit, or to apply your own variant. Can we collapse that into one catch-all question?
C: [Writes it as the closing question, naming spellcheck, predictive text, model naming, NDAs and file naming explicitly, so that a reader arriving with one of those questions can see it has been considered rather than overlooked.]
One sentence is mine and worth your judgement: the only thing against the spirit of it would be settling such a question quietly, in a way you would rather a reader didn’t know about. Without it the answer is a shrug, and a shrug invites the reading that Kladde means nothing in particular. But it does introduce a “spirit of the thing”, which a document with no authority can’t enforce.
[J keeps it.]
Moulding
Mostly J leaving comments on the draft.
J (on “whose manifesto is the model for this one”): the tone of which inspired
C: [Changes it.] The weaker claim is the more accurate one. Dogme’s manifesto isn’t Kladde’s model in any structural sense — no numbered rules, no craft constraints, no vow — and claiming it as the model invites a comparison Kladde loses.
J (on “Any such number would be arbitrary, unverifiable and easy to game”): We believe that in a real-world collaboration this is impossible to measure in any fruitful manner, and the number would be impossible to interpret meaningfully.
C: [Takes it verbatim.] Yours makes the better argument. “Arbitrary, unverifiable and easy to game” attacks the number’s reliability, which concedes that a trustworthy number would be worth having, and invites someone to go and invent one. Yours says the quantity doesn’t exist to be measured.
That’s also the first “we” in a document otherwise speaking as “I”.
J: Yeah, for tone, go from “I” to “we” throughout, matching the manifesto.
C: [Converts the answers, and leaves the questions alone.] The “I” in the questions is the reader’s — “Why do I need this?”, “How do I mark a work?” — so an FAQ properly has two voices. Converting the questions would have had the manifesto interrogating itself.
J (on “provenance”): Avoid provenance, it’s the language of forensic evidence again.
C: [Replaces it with “how it was made”, and checks that it was the only instance. The two uses of “forensic” are J’s own, both disclaiming the standard rather than invoking it, and stay.]
J (on “Then say that”): It’s up to you whether you believe Kladde adds value, for any given work. In this case, if you do choose to use Kladde, simply say what you did.
C: Yours fixes something mine got wrong. “Then say that” answered as though the obligation were settled and only the wording remained, which assumes the reader has already signed up.
J (on “A writer may publish some work as Kladde and some not”): Add, however, that by publishing most work as Kladde but silently omitting it in others, you are strongly implying that the non-Kladde work is AI-free. The honest move is to disclose when it is not.
C: [Adds it.] Once a writer marks work habitually, absence stops being neutral and starts being a statement. But this is the first thing in any of these documents that asks something of work outside Kladde. It could warn instead of instruct: …whether you meant to or not.
J: Yes change it to not instruct.
Late additions
J: Another FAQ might be “Is this only for online publishing?” Not at all — it may not be practical to print Kladde workings, but you can print a URL linking to the workings. Consider link shorteners or QR codes, the usual tricks used to bridge the physical world to the Web.
C: [Adds it before the reprint question, and cuts “Nobody expects a print journal to typeset a transcript” from that answer, which now repeated the point.]
J: Add a “Where can I see examples” question and link to john.hartnup.uk/fiction.
C: [Adds it second, after the marking question.] The HTTPS version of that address redirects to plain HTTP, so the certificate may not be provisioned yet.
J: https should now work, so use that.
C: [Changes it, then checks twice, the second time with a cache-busting query string, and finds it still redirecting to HTTP. Writes https anyway, since that is where it is heading, and says the check failed rather than reporting it as confirmed.]
J (on “Our own fiction is at”): The first known examples are at
C: [Changes it.] Less proprietorial, and it assumes there will be others. Small hostage to fortune: “known” implies somebody is keeping track, when really it’s known to us.
J: Do known to us.
The FAQ was written on 18 September 2026, in about ninety minutes. Six of its questions are J’s, with their answers substantially his: the sole-author question, not wanting to take part, the effort involved, transcript length, falsification and whether readers will bother. Three more are his additions — printing, examples, and the catch-all, whose answer is his and whose drafting is C’s. The remaining ten were proposed by C from decisions already taken, and cut back by J comment by comment.
The definition of the mark is J’s, prompted by C noticing the manifesto used a term it never explained. The instruction to speak as “we”, the removal of “provenance”, and the refusal to turn a warning into a rule are all J’s, and all three moved the document away from how C had first written it.