Richard Bellingham @skimble.bsky.social · 16/09/2026That is how I've always thought about it. I might write a plot involving some existing npcs, character books etc. but what happens from there unfolds naturally. Sometimes it's a literal plot, as in, 'here is the NPC's plan, let's see what happens." 020
Richard Bellingham @skimble.bsky.social · 01/09/2026"You can totally trust our engines not to hallucinate or make things up. They are called 'Fable' and 'Mythos'." 120
Richard Bellingham @skimble.bsky.social · 21/08/2026... But as a tool used judiciously by someone who already knows what they're doing and not relied on without verification, it's not entirely useless. I did get one really useful thing out of it recently; replacing a table in a PDF. We don't have PDF editing software, so that saved a lot of time. 010
Richard Bellingham @skimble.bsky.social · 21/08/2026That makes sense.. I'm only using it at work to translate emails or to get the gist of insurance documents. I use it to proofread emails and it is good at spotting inconsistencies and typos; inferior to a second pair of eyes but that's not feasible. I think the tech has immense capacity for harm... 110
Richard Bellingham @skimble.bsky.social · 21/08/2026Do you mean using it to generate text/images/music, or are you including all uses of LLMs including translation, document comparison, and proof reading? I won't use them for generation, but have found some use analytically as long as I don't abdicate final responsibility. 210
Richard Bellingham @skimble.bsky.social · 17/08/2026I just finished @wiswell.bsky.social's "The Dragon Has Some Complaints". It was good! You should read it if you like dragons. So many dragons. And romance. And comedy. And drama. But no Bad Dragons in here. 2193
Richard Bellingham @skimble.bsky.social · 17/08/2026And he's voiced by the Twink Overlord, Jack Quaid! It's a pretty good show. 010
Richard Bellingham @skimble.bsky.social · 17/08/2026And then a Stephen King reference just a few pages later! I wouldn't have been able to resist that, either. 000
Richard Bellingham @skimble.bsky.social · 17/08/2026@wiswell.bsky.social I'm not sure if I want to punch you or applaud you for this one! 100
Richard Bellingham @skimble.bsky.social · 11/08/2026I mean, I'm a human, he seems pretty down on the whole concept of humans... 010
Richard Bellingham @skimble.bsky.social · 09/08/2026Unironically one of my favourite films. "It's got a SPINE, doesn't it?" 020
Richard Bellingham @skimble.bsky.social · 17/07/2026Allegedly it's only 700ish REAL pages, but the large print for my presbyopic reading without reading glasses makes it longer! 000
Richard Bellingham @skimble.bsky.social · 17/07/2026I started a 2,000 page (at least, on my Kobo, not sure how many paper pages that is in reality) book in between, so now I have to read that first, bah! Looking forward to this immensely, though. 210
Richard Bellingham @skimble.bsky.social · 07/07/2026@aptshadow.bsky.social "...laughs again, despairingly, and tells him." I want to punch you right now. :) 010
Richard Bellingham @skimble.bsky.social · 04/07/2026Yeah, I really enjoyed it too. I think Superman was the better film as a whole, but this was solidly entertaining and had a rich emotional core surrounded by fun action sequences. References to John Wick are apt with regard to canine relationship as in both movies the dog is a grief support animal. 020
Richard Bellingham @skimble.bsky.social · 01/07/2026I should have read the page... There are definitions but no examples for both Mary and Marty because both types are just too prone to flame bait! The Gary Stu type definitions are worth a laugh though. 010
Richard Bellingham @skimble.bsky.social · 01/07/2026There is the male equivalent of a Gary Stu/Marty Stu. I think it has to be a really egregious example before the term is applied so I'm not disputing the double standard. Tvtropes has some examples. tvtropes.org/pmwiki/pmwik...tvtropes.orgMarty Stu - TV TropesMary Sue, and the plot bias indicative of it, is hardly limited to those of a female persuasion. For probably just as long (if not even longer) as there has been Mary Sue, so too has there been Marty ... 130
Richard Bellingham @skimble.bsky.social · 23/06/2026Released in the UK 3 days later? What treachery is this? 000
Richard Bellingham @skimble.bsky.social · 14/06/2026I know Fate isn't quite crunchy enough for you but this is also the space occupied by its stunts and aspects. It can have its own issues with decision paralysis on coming up with personalised stunts. I think historically tons of options/crunch were provided as it makes for marketable products. 010
Richard Bellingham @skimble.bsky.social · 07/06/2026I also find it useful to feed it contractual clauses and ask if to interpret them. It gives a baseline idiot's interpretation that I can use as evidence it could be read a certain way. I don't use it to interpret for me, just get a second opinion on the possible interpretation. 000
Richard Bellingham @skimble.bsky.social · 07/06/2026The main use I've found for CoPilot at work is suggesting edits to work emails to reduce the cognitive load and make them easier to action. I don't let it write for me, but it's pretty good at feedback for the corporate communication space. 110
Richard Bellingham @skimble.bsky.social · 06/06/2026If you liked It Follows you might like Insidious. It's a bloodless supernatural horror film. Jump scares but also genuine tension and sensible protagonists. The sequels are decent as well. 030
Richard Bellingham @skimble.bsky.social · 06/06/2026There are a lot of horror films that aren't about degrading their audience or even their characters. That's mainly a thing with slasher or torture porn horror, which hasn't been much of a thing in the 2020s. With the exception of the Terrifier films. 110
Richard Bellingham @skimble.bsky.social · 06/06/2026Obsession was a fucking great movie and it deserves to do super well. I'm a horror aficionado and it's one of the few films to make me feel genuinely tense in months. Fantastic performances, as well. Highly recommended. 010
Richard Bellingham @skimble.bsky.social · 06/06/2026Does this count young nerds who don't like sports and get very upset when the entire TV schedule is taken over by sports and I can't watch my Knight Rider? Asking for a friend. 120
Richard Bellingham @skimble.bsky.social · 04/06/2026Shibboleths of humanity in an increasingly inhuman world. 020
Richard Bellingham @skimble.bsky.social · 04/06/2026I predict a movement towards baroque language, pettifoggery, and sesquipedalianism in human writing along with deliberate use of whimsical and unusual verbiage as a significator of humanity as a conscious rejection of the surface gloss and low grade level readability of LLM output. 210
Richard Bellingham @skimble.bsky.social · 04/06/2026Yes, humans in aggregate do use those linguistic structures... Expected as LLMs are trained on human writing and use probabilistic models to generate output. An agglomeration of multiple of them in the same short piece is suggestive. As you say, it's impossible to prove. So we're back to vibes. 100
Richard Bellingham @skimble.bsky.social · 04/06/2026Rigorous, organised structure with ruthless conciseness Wrapping everything up in a philosophical bow even if that's not really congruous with the piece High school essay style structure These are all signals, but not completely decisive. 000
Richard Bellingham @skimble.bsky.social · 04/06/2026I listed several of them as also identified by Gemini, but you rejected them. "This isn't just x, it's Y" "Just a quiet argument" Use of colons where humans would probably use commas or new sentences "And that's exactly what..." Use of bold text or italics for emphasis, em dashes, emojis (weaker) 200
Richard Bellingham @skimble.bsky.social · 04/06/2026One of the most corrosive aspects of LLM use is that there's always going to be an element of doubt as to whether something was created with intentionality by a human or procedurally generated. This is happening in the art world now with people being accused of using genAI even if they have not. 010
Richard Bellingham @skimble.bsky.social · 04/06/2026There's never going to be a concrete set of criteria that will always give the game away, I'm afraid, so intuition is generally the best you can do. Common signifiers like 'it's not just x it's y', use of em-dashes, use of emojis and colons, and bullet point structuring are all used by humans too. 200
Richard Bellingham @skimble.bsky.social · 04/06/2026There is no single tell that an LLM has that couldn't also be expressed by a bland human writer, but you do start to get a feel for it when you've seen enough examples. The more of the common structures and phrasing you find in a given piece, the more likely it was generated by an LLM. 100
Richard Bellingham @skimble.bsky.social · 04/06/2026The problem is that LLMs are trained on vast swathes of human writing, so they are pretty good at cobbling together generic human writing without any artistic merit or human intentionality. They are particularly good at corporate content and puff journalism due to their training. 110
Richard Bellingham @skimble.bsky.social · 03/06/2026You know, like, it's not just a soundbite polished to within an inch of its life. It's human, conversational. And that's what's important. 000
Richard Bellingham @skimble.bsky.social · 03/06/2026Oh yeah, definitely. I don't think I've ever seen a human write that... much more usual to use a comma, like: "Credit where it's due, I've been getting tips from CoPilot on how to write short, personality-less work emails that are as efficient as possible and it's... pretty good at that." 100
Richard Bellingham @skimble.bsky.social · 03/06/2026Pfft, look at George R R Martin over here. :P 020
Richard Bellingham @skimble.bsky.social · 03/06/2026I can't really put my finger on it. Much in the same way I write by instinct, I think I have an instinct for what feels artificially generated in writing. The biggest tell for me is usually stuff like, "It's not just X, it's Y."or "That's exactly what you SHOULD be doing." and so on. There's a vibe. 100
Richard Bellingham @skimble.bsky.social · 03/06/2026@craigschaeferbooks.com Jeebus sexy rat girls, I think Rats might be your best book yet. Though, I mean... I haven't finished it. There's a CHANCE you could botch the ending. When're the fourth and fifth books coming out? I've read two and two thirds books in quick succession and I need more. 010
Richard Bellingham @skimble.bsky.social · 03/06/2026So I had to correct it on that one... Here's the whole Gemini thread: gemini.google.com/share/160eeb...gemini.google.comGemini - direct access to Google AICreated with Gemini 000
Richard Bellingham @skimble.bsky.social · 03/06/2026I think it was the "It's a quiet argument..." that just immediately triggered my LLM detector. Gemini admitted it has hardcoded rails from Google to prevent it from using the same rhetorical devices and structure. At first it said this was LLM generated because it was set in 2026 and we're in 2024 110
Richard Bellingham @skimble.bsky.social · 03/06/20263. Structural Pacing The article follows a rigid, algorithmic essay structure: The Hook: The historical anecdote The Reveal: What the tech is. The Mechanics: How it works. The List: Potential applications. The Philosophy: What this means for material science. The Bow: A tidy, thematic conclusion. 100
Richard Bellingham @skimble.bsky.social · 03/06/2026It ends with a summarizing moral: "The Y-Zipper's story is, among other things, a quiet argument for why we should be much slower to reject ideas..." 100
Richard Bellingham @skimble.bsky.social · 03/06/20262. The "Moral of the Story" Conclusion LLMs are notoriously programmed to end articles with a sweeping, philosophical takeaway or a neat moral wrap-up. It pivots to a broad philosophical point: "That's a fundamentally different way of thinking about how we build things." 100
