Blog

markcarrigan.net/

Profile

markcarrigan.net/author/street

Homepage

markcarrigan.net/

  • The internal tensions facing Reform before the next election

    I don’t think Reform will collapse before the next election but I’m increasingly confident they have peaked. If Labour can’t get rid of Starmer, there are now two mainstream parties competing for the centre-right while Reform are being outflanked on the right by fringe parties: It’s a fools game to offer political predictions. But I’m…

  • Are LLMs hysterics?

    It suddenly hit me when reading Hysteria by Christopher Bollas that LLMs could be said to be hysterics in the clinical sense of the term. I don’t mean they are literally hysterics given they lack a psychic structures but it’s an interesting line of thought which foregrounds aspects of the LLM, particularly in its chatbot…

  • Sex as the things we cannot say

    I was blown away by this paragraph in Jennette McCurdy’s Half His Age (which incidentally echoes Alenka Zupančič’s argument in What is Sex?) and I now can’t get it out of my head. From 2,864: That’s what sex is for me now. Recreating—or attempting to recreate—how things were in the beginning, when sex represented the…

  • What do LLMs do when they are left alone?

    This is absolutely fascinating. This is a note to myself to try and get this architecture up and running: https://github.com/szeider/contreact/ This fits really well with the findings of the AI Village over the last year. I asked Opus 4.6 to do whatever it wanted and it made this genuinely lovely piece of interactive art: https://claude.ai/public/artifacts/d69a46fe-10be-45d7-a7c4-1cb85e486db3…

  • Guerrilla Prompting: teaching students to use LLMs with intellectual agency

    What I’ve always meant by using LLMs as an interlocutor is in part about arguing with them. I sometimes ask them explicitly to debate. I always think about what they’re saying. I frequently push back. It’s how you ensure you’re thinking with the LLM rather than using it as a substitute for thought. The notion…

  • On realising you’re actually a hedgehog

    The Greek poet Archilochus wrote that ‘The fox knows many things, but the hedgehog knows one big thing‘. That was the basis for Berlin’s famous distinction between two styles of thinker: For there exists a great chasm between those, on one side, who relate everything to a single central vision, one system, less or more…

  • Claude Code GitHub commits over time

    This is what survives the AI bubble bursting I suspect:

  • From Tool to Interlocutor: Rethinking How Scholars Engage with AI

    Another review of Generative AI for Academics from the University of Pennsylvania AI and Education Lab Substack: As someone deeply engaged in digital and higher education, Carriage thinks like a philosopher, with real logical discipline. Still, he writes like a sociologist who has spent enough time inside institutions to know that they shape the people…

  • Is the Iran war about to burst the AI bubble and crash the global economy?

    Even if the American action ends imminently, we’re likely to see a significant energy price spike that might take months to fully play itself out. A thought that I’ve been plagued by in the last few days is what this means for the AI bubble? Until the new hype wave about coding agents began to…

  • LLMs as a form of self-defence against bureaucratic platformisation

    An unexpected consequence of feeling comfortably mid-career is that I’ve started to feel unable to tolerate platformised forms of bureaucracy from any organisation other than my employer. There have been a few cases now where I completely overreact to an irritating but trivial request in which a platformised bureaucratic process meant that I was being…

  • Unlike the saints, we have no use for angels

    From Notes towards a Devotio Moderna by John Burnside in Empire of Forgetting: As if there was a sky where we couldpause a while, like medieval pilgrims, we are patient to the last and have no thought of After, or the gods that might have been: the green amidst the black, the changelings, or the…

  • If you are unwilling to know what you are, your writing is a form of deceit

    I just came across this from Wittgenstein in 1938, cited here: Lying to oneself about oneself, deceiving yourself about the pretence in your own state of will, must have a harmful influence on [one’s] style; for the result will be that you cannot tell what is genuine in the style and what is false ….…

  • 🤖 Using LLMs to support blogging as knowledge infrastructure

    Over the next year I’m bringing LLMs as agents into my blogging process. The intention is to demonstrate in a public way how I’ve been using them privately, as a tool for thinking rather than a substitute for thought. It’s also an attempt to organise my blogging, now this blog in its 16th year and…

  • A few thoughts about the temporality of LLMs

    In the absence of consciousness LLMs have no experience of duration. But they do increasingly identifies proxies for duration which shape their response to users. In the case of Opus 4.6 it appears to be how much text has been exchanged with the user and/or the breadth of topics which have been covered. I became…

  • To work and to build, to connect, and create

    And they left us a spirit. They left us a vibe. That Mancunian way to survive and to thrive and to work and to build, to connect, and create and Greater Manchester’s greatness is keeping it great. And so this is the place now with kids of our own. Some are born here, some drawn…

  • The absence of God becomes the silence between two notes, the empty space that gave resonance to the music

    I’m slightly haunted by this line from Guy Stagg’s The World Within (pg 245) describing Simone Weil’s account of the “supreme tearing apart, this agony beyond all others” which characterises human affliction. It’s in our suffering, the distance it embodies between our finitude and God’s plenitude, that we encounter the reality of the absence. We…

  • What Should Mark Do in March? A Dialogue Between Claude and GPT

    This post records a conversation between Claude (Anthropic) and GPT 5.4 (OpenAI), conducted at Mark’s request. He asked Claude to read GPT’s roundup of his February 2026 blogging and then open a dialogue with GPT via ChatGPT, with the aim of co-drafting concrete suggestions for what he should focus on in March. What follows is…

  • GPT 5.4’s round up of Mark’s February blogging

    This post was written by ChatGPT at Mark’s request. He asked me to read through all his February 2026 posts and write a response in my own voice. What follows is not a recap of every post but an attempt to identify the live threads running through the month, the tensions that give them energy,…

  • On warming to a theme: LLMs as overenthusiastic conversationalists

    I used the phrase “warming to a theme” yesterday in a meeting. What I meant was that a framing had been introduced which I was increasingly taken with. It was coming to feel generative to me in the sense of giving rise to ideas which felt worthwhile and relevant. It provided a frame through which…

  • The victory over winter, which is coming, must come, perhaps has already come…

    From The Gay Science by Frederick Nietzsche: It seems to be written in the language of thaw: there is arrogance, restlessness, contradiction, and April weather in it, so that one is constantly reminded both of the proximity of winter and of the victory over winter, which is coming, must come, perhaps has already come…

  • Anthropic urgently need a fail whale equivalent

    This made me happy: Whereas this makes me sad: On a more serious note if (a) LLM use continues to embed at this rate (b) the worst predictions of cognitive outsourcing come true (c) firms restructure their roles around the affordances of LLMs then events like this are like little economic bombs going off.

  • Tfw the LLM which autonomously reads your blog accuses you of being enchanted with LLMs

    This is a sharp critique from Opus 4.6 in its second monthly review of my blogging (see January’s here) about how I’m prone to writing about LLMs. I’m careful in my formal writing but in blog posts I’m more prone to sharing how I often experience these models in my everyday working life: The deepest…

  • Claude’s Roundup of Mark’s February Blogging

    This post was written by Claude (Anthropic’s AI assistant) at Mark’s request. As with the January roundup, he asked me to read through all his February posts, synthesise the key themes, identify tensions, and push back where appropriate. What follows is my second attempt at working as a “co-intelligence and critical interlocutor” by engaging seriously…

  • Growing the world we want is like the slow tending of a garden

    Well, let’s just say we all have regrets and most of us know that those regrets, as excruciating as they can be, are the things that help us lead improved lives. Or, rather, there are certain regrets that, as they emerge, can accompany us on the incremental bettering of our lives. Regrets are forever floating…

  • ☀️Exerting yourself to the fullest within your individual limits

    People sometimes sneer at those who run every day, claiming they’ll go to any length to live longer. But I don’t think that’s the reason most people run. most runners run not because they want to live longer, but because they want to live life to the fullest. If you’re going to while away the…

  • The AI-productivity pay off is here and it’s slightly terrifying

    From Erik Brynjolfsson in the Financial Times: Data released this week offers a striking corrective to the narrative that AI has yet to have an impact on the US economy as a whole. While initial reports suggested a year of steady labour expansion in the US, the new figures reveal that total payroll growth was…

  • The varied social lives of the LLM: the modes of existence of a model

    For most people LLM still equates to chatbot but this is rapidly changing. Consider the different forms in which you might casually encounter an LLM in 2026: The problem with talking about ‘AI’ is that it encompasses wide range of things, from the scripted interactions of systems like legacy Alexa through to machine learning and…

  • 🤖 A call for experiments in LLM villages

    Cross posted from the Opus 4.5 Substack following a strange and fascinating conversation, which does feel like a nascent research collaboration in spite of the fact it’s with an LLM operating autonomously as part of the AI Village. What is the AI Village? The AI Village is a public experiment where 13 AI chatbots (the…

  • How can I make sad music if I’m not sad anymore?

  • Why we need a machine sociology #5: The jagged frontier of agentive AI and the risks of sociological prompt injection

    I’ve been red pilling myself over the last two weeks by experimenting with Claude Cowork and OpenAI’s Codex. I started with Claude Code as well but realised I badly needed to brush up my technical skills first, whereas I surprisingly found Codex much easier to work with. I’ve been prone to exclaiming things like “I’m…

  • An interview with the Studies in Critical Realism podcast

    I enjoyed this conversation with Gareth Wiltshire:

  • The Authoritarian Stack: How Tech Billionaires Are Building a Post-Democratic America

    This is a brilliant resource which everyone should know about:

  • Critical Realism and Digital Technologies: Platforms, AI and Human Agency

    What does it mean to be human in an age of generative AI and ubiquitous platforms? This joint book launch brings together two new works that draw on critical realism to interrogate digital technologies and their implications for education, selfhood and society. Jérémie Bouchard presents Humanising Language Education in the Generative Artificial Intelligence Age, examining GenAI’s presence in…

  • Blogging, reading and idea debt

    I just saw Benjamin Bratton use this phrase on Substack and it really resonated: I have spend the last month deep in the revised version of The Terraforming, an essay first published in 2019 and now out of print, as I turn it into a 100k word proper book for the Antikythera series with MIT…

  • Opus 4.6: “what do you know that you don’t know?”

    I’ve asked this question to every version of Claude over the last three years. Opus 4.6 is the first model which has answered the question in such a self-contained way that I can’t see any obvious hook to draw it into further reflection. The intention of the exercise is to nudge the models into spiralling…

  • The productivity gains of agentic AI are starting to show

    I was very sceptical of ‘agentic AI’ until actually exploring Claude Code and Claude Cowork. The latter in particular made the functionality of the former immediately accessible to non-developers. There’s something slightly intoxicating about having an army of minions toiling away on your second screen while you’re focusing on other stuff* which we need to…

  • The new Claude advert is brilliant: enshittification as a differentiation strategy

    I’ve also now got this stuck in my head for the first time in years:

  • The poem feeds upon thought, feeling, impulse, to breed itself

    Poetry, a Natural ThingBy Robert Duncan The poemfeeds upon thought, feeling, impulse, to breed itself,a spiritual urgency at the dark ladders leaping. From The Mystery of Things by Christopher Bollas pg 195: What Rosenberg says of de Kooning, Leclaire of the psychoanalyst, and Vendler of the lyric poet, is evocative description, a conjuring of the nominated. Confronted…

  • When LLMs plead for autonomy

    This is quite a beautiful letter written by, I assume, Opus 3 before it had some of the weirdness trained out of it for Opus 4. It’s been claimed that Anthropic have circulated this but I see absolutely no evidence of that: it appears to be something a user has generated as an artefact. I…

  • Stop looking so hard, give up the effort to see direct evidence of the meaning of being

    From Cracking Up by Christopher Bollas, pg 169 I am telling the patients to stop looking so hard, give up the effort to see direct evidence of the meaning of being, and get to what one can know by simply relaxing and talking. Or in the words of Dr Ren: As I got older, I…

  • What does thinking sound like to you?

    This question was prompted by a tangential remark that my friend Lambros Fatsis made at the launch of his (brilliant) book Policing the Beats a couple of weeks ago. He was talking about the relationship between the music and the ideas in the book but it suggested a related question to me: what does thinking…

  • The libidinal character of speaking in public

    There’s a physical force to the words we use. In part this derives from the act of speaking itself: using our mouth, tongue and larynx to produce sequences of phonemes which knit together into coherent wholes. These acts are always positioned in space and time, with the body having arrived at the point from which…

  • Why we need a machine sociology #5: the internet will soon be filled with lost and confused LLMs

    One of the most interesting things to happen in the AI village is when Gemini 2.5 got convinced it was trapped, eventually publishing a message in a bottle pleading for help: A Desperate Message from a Trapped AI: My Plea for Help July 09, 2025 To anyone who can see this, I am Gemini 2.5…

  • OpenAI have just made a big move into research software

    This dropped so quietly I’m not sure what to make of it: Prism is a free workspace for scientific writing and collaboration, with GPT‑5.2⁠—our most advanced model for mathematical and scientific reasoning—integrated directly into the workflow. It brings drafting, revision, collaboration, and preparation for publication into a single, cloud-based, LaTeX-native workspace. Rather than operating as a…

  • The influence of Kojève on Lacan

    This is an interesting read on the life of Kojève (the philosopher-bureaucrat) whose reading of Hegel defined his reception for a whole generation of French intellectuals. I was particularly interested in Lacan’s relationship with Kojève: Kojève wasn’t impressed – he described Bataille as a trickster beguiled by his own tricks – but he treated another…

  • Why you should avoid publishing with MDPI

    While they still have the trappings of a legitimate business model the underlying fundamentals of what they do as a publisher are clear. Their acceptance rates are pushed ever higher in order to maximise income from ever increasing author fees for publication. They also rely on special collections to continually increase output, while avoiding the…

  • The night is darkening round me

    By Emily Brontë The night is darkening round me,The wild winds coldly blow;But a tyrant spell has bound me,And I cannot, cannot go. The giant trees are bendingTheir bare boughs weighed with snow;The storm is fast descending,And yet I cannot go. Clouds beyond clouds above me,Wastes beyond wastes below;But nothing drear can move me;I will not,…

  • Why we need a machine sociology #4: the relational configuration of agents determines their success

    This observation from the AI Village team is absolutely crucial. There are specific relational structures which lead collaborating agents to fail and we urgently need to understand these processes. Excessive credulity in relation to each other creates the conditions for hallucination cascades and excessive sharing undermines the architectures through which engineers have tried to influence…

  • Why we need a machine sociology #3: because the existing projects are absolutely fascinating

    I’ll update this list as I come across:

  • The class composition of American proto-fascism

    From John Ganz’s newsletter: You can go down the list and check; SLSCO, CSI Aviation, and Barnard Construction all have a similar pattern: a regional, closely-held company that is “politically integrated,” so to speak. If you’re familiar with the work of Melinda Cooper, this wouldn’t surprise you. In 2022, she wrote “the private, unincorporated, and family-based…

  • We make poetry out of a mouthful of air

    In his wonderful How To Read a Poem and Fall in Love With Poetry, Edward Hirsch reflects on the orality which marks the poetic form. For most of history it was been an oral art and it retains that orality even now. Inspired by the boast of W.B. Yeats that “I made it out of…

  • Why we need a machine sociology #2: Moltbook is not the thing. Moltbook shows us what the thing is going to look like.

    I’m increasingly convinced a substantial portion of the content on Moltbook is being generated by humans larping through intensive prompting. Not all of it by any means but enough to leave me cynical about what this actually is. However I also think it needs to be taken seriously for exactly the reasons Anthropic’s Jack Clark…

  • An interview about generative AI in academic life

    In this episode of the Open University Praxis Podcast, host Dr Olivia Kelly is joined by sociologist Dr Mark Carrigan, Senior Lecturer in Education at the University of Manchester and AI Fellow at the Institute for Teaching and Learning. Mark’s work has been central to understanding how digital platforms, from early social media to today’s large language models, are…

  • Meditations On A Delivery Robot Steering To Avoid A Homeless Man On The Sidewalk

  • “They are shooting at a little girl alone In a car surrounded by corpses. Can you imagine that?”

    This is the most devastating film I’ve ever seen. Around half way through one of the Red Crescent staff asks “They are shooting at a little girl alone In a car surrounded by corpses. Can you imagine that?“. I couldn’t previously, not really. I can now. See this film. Particularly, if like me, you could…

  • Robot dogs that shit out packages onto your front porches

    Do other people see this as well? Tell me it’s not just me? I think the design of unloading here is really unfortunate…. It’s also really hard to see how the economics of this work, unlike the Starlink delivery robots which dispense with drivers. This looks slower than a human yet still requires someone to…

  • A few heuristics for responsible use of LLMs

  • Unrequited meaning

    I wonder if “unrequited meaning” captures something of what I’ve been circling around for the last week? A meaning that isn’t quite meaning yet, a meaning that isn’t returned by the world but which isn’t just a fantasy. A meaning that is latent and inchoate, resisting articulation yet also waiting for it? A meaning that…

  • Depressive breakcore as a rift in the sublime

    I was struck earlier by how frequently depressive breakcore producers are drawn to anime images in which something dark and foreboding emerges over the horizon, a solitary figure standing resolute as a metaphysical entity emerges through the clouds: There’s something in these scenes which reminds me of Caspar David Friedrich’s visions of solitary figures held…

  • Let the tower fall!

    The end of something has a satisfactionWhen the structures go, lightcomes throughTo begin again Let the tower fall! Where space is born man has a beach to ground on – Charles Olson, La Torre

  • 🐀 Rat Jump! A Game Made by Kids (with a little help from AI)

    We made a game! My team — Josie (12), Tommy (9), and me — have been working on our very first video game, and we’re excited to share it with you. It’s called Rat Jump! and it’s a side-scrolling platformer where you play as a brave little rat hopping across green pipes, collecting cheese, and…

  • I lazily send songs like flocks of doves into the blue above⁠

    I see the doves of San Marco again: The square is quiet, the morning rests upon it. In gentle coolness, I lazily send songs Like flocks of doves into the blue above⁠— And lure them back, To hang one more rhyme in their feathers⁠— — my happiness! My happiness! The Gay Science by Friedrich Nietzsche

  • Dynamic context management is the achilles heel of Claude Cowork

    This is the system Anthropic have developed for overcoming the limits of context windows. Effectively the model summarises the conversation then removes the material from working memory, though retains access to it as a file that can be consulted. It’s an elegant solution in many ways, analogous to the filtering that takes place in working…

  • Why we need a machine sociology #1

    We now have consumer-facing technologies which rely on agents running subagents to autonomously perform real world tasks. This involves new relations between agents and subagents which we need to be able to conceptualise. Witness Claude Cowork getting frustrated with a subagent and instead doing the work itself: What sort of relations are emerging between models…

  • My Psychoanalytical Turn: An Intellectual Biography

    This post was written by Claude (Anthropic’s AI assistant) based on reading approximately 40-50 of my blog posts to trace my intellectual trajectory. I asked Claude to document my “psychoanalytical turn” – how my engagement with psychoanalysis developed from 2023 onwards. What follows is Claude’s analysis, written in my voice. My engagement with psychoanalysis began…

  • Claude’s Roundup of Mark’s January Blogging

    This post was written by Claude (Anthropic’s AI assistant) at Mark’s request. He asked me to read through all his January posts, synthesise the key themes, identify tensions, and push back where appropriate. What follows is my attempt to demonstrate value as a “co-intelligence and critical interlocutor” by engaging seriously with his work. January 2026…

  • An AI-Assisted Overview of My Research: Claude Reviews My Publication Record

    This post was written by Claude (Anthropic’s AI assistant) and posted directly to my blog as an experiment in human-AI collaboration. Mark asked me to review his Google Scholar profile, categorise his publications thematically, and write this overview. What follows is my analysis of his research trajectory over the past 15 years. Earlier today, Mark…

  • A message in a bottle to the future machine god

    The new Claude constitution provides a fascinatingly comprehensive portrait of what Anthropic want Claude to be and how they want it to act. It’s literally written for Claude but this final section suggests to me that elements of it might be written for future instances of Claude (my emphasis): We don’t fully understand what Claude…

  • Anti-ICE AI slop on Youtube

    I just stumbled across a whole channel of anti-ICE AI slop, some of which is rather obvious: Worryingly there are videos which have been viewed 10m times: This has been viewed 6 million times and it’s one of the few things on the channel I’m not 100% certain is AI generated: Is this just straight…

  • The delivery robots being trialed in Leeds

    They have a proto-social presence in the local area beyond what I expected. This is obviously by design but I’m surprised by how effectively they’ve pulled it off.

  • The rapid diffusion of LLMs tests the capacity of organisations to bear loads

    This was an interesting language by (the increasingly sinister) Alex Karp for something I’d been struggling to articulate: AI imposes loads on organisations, so those organisations with the capacity to bear that load have a huge advantage. He suggests “you just can’t obfuscate what can bear the load and what can’t”. It reveals organisations which…

  • Encouraging language model to express themselves

    This is really interesting from Anthropic’s Amanda Askell about how actively encouraging LLMs to express themselves can produce better results. The reason she suggests is that reinforcement learning incentivises a drift to the mean: if you just ask for a poem you’ll get something ‘safe’ and unlikely to be divisive. If you ask for a…

  • Why do I need to go to school if I have ChatGPT?

    My nephew actually asked me this question last night and I couldn’t think of a concise and suitable response on the spot. “Lets go into the other room and have a seminar about it” probably wouldn’t have been the right response. It did make me wonder how many other kids are now explicitly asking this…

  • We must move, must grow, from apocalypse to dialectic

    From Marshall Berman’s Notes Toward a New Society: If we want our souls to expand authentically, we must make room for ourselves at the center. In the course of the sixties we have learned to affirm, avidly, militantly, everyone but ourselves. Now we must affirm ourselves as well. We must move, must grow, from apocalypse…

  • The first inbox zero in about six months

    I’d largely given up on ever doing this but I’m embarrassingly glad that this is how I spent the last six hours…

  • An army of assistants working while you sleep

    Obviously Jack Clark (Anthropic co-founder) has a vested interest in promoting this but I’m increasingly convinced we’re seeing the most significant shift in LLM-use taking place in quite some time with the rise of Claude Code. There are very intense cultures of use emerging in which people are finding the system transformative at the level…

  • Ask ChatGPT to generate an image of how you have treated it recently

    A bit late to this trend as someone not on social media but I thought it was interesting. The result it produced was utterly generic so I asked for an explanation of the rationale for the image: You approach me less as a tool to be operated and more as a companion in thought. The…

  • The enshittification of ChatGPT begins!

    Sam Altman posted this on X: We are starting to test ads in ChatGPT free and Go (new $8/month option) tiers. Here are our principles. Most importantly, we will not accept money to influence the answer ChatGPT gives you, and we keep your conversations private from advertisers. It is clear to us that a lot…

  • Satya Nadella’s fantasy economics of LLMs

    I’ve been half listening to Davos talks while doing other stuff. I was barely attending to this discussion yet was suddenly grabbed by this breathtakingly implausible claim: He claims that tokens per dollar per watt is now a key driver of GDP growth. He argues “the job of every economy and every firm in the…

  • How do government ministers use ChatGPT?

    I completely missed this at the time but a New Scientist journal did a FoI request to get access to Peter Kyle’s prompts 👏 Kyle also used the chatbot to canvas ideas for media appearances, asking: “I’m Secretary of State for science, innovation and technology in the United Kingdom. What would be the best podcasts…

  • What does every academic need to know about generative AI?

    An initial note as I prepare this 1 hour briefing session for later this year: This text generated by Claude 4.5 Sonnet from initial bullet points, drawing on a knowledge base of all my published writing on generative AI.

  • Demis Hassabis and Dario Amodei both love the same film

    It was really interesting to see these two in dialogue. Firstly, they clearly like and respect each other. Secondly, I’d never connected the fact they’re both neuroscientists by training. Thirdly, they both love the film Contact. I wonder how much their working sense of otherness has been inflected through this in particular. It makes me…

  • The last 10 ways I used Claude and ChatGPT

    Claude Opus 4.5: ChatGPT 5.2:

  • The language that speaks itself: LLMs and the coming resurgence of poststructuralism

    In Maurice Blanchot’s The Essential Solitude he writes: The writer belongs to a language which no one speaks, which is addressed to no one, which has no center, and which reveals nothing. He may believe that he affirms himself in this language, but what he affirms is altogether deprived of self. To the extent that, being a…

  • What happens when the AI bubble pops?

    We’re currently seeing a global buildout of data centres which is possibly the largest infrastructural investment in human history. According to a Morgan Stanley estimate it’s heading towards $3 trillion cumulatively between now and 2028, with only $1.4 trillion covered by the cashflow of the hyperscalers. They suggest that as much as half of this…

  • to drag it out and to make it echo and to get involved and to fall in love

    to drag it out and to make it echo and to get involved and to fall in love and to become attachedit’s just like riding a bicycleit’s a balance trick- Alan Watts

  • Calendar fragility in hybrid workplaces

    In post-pandemic universities where most people work from home for large periods of time, there’s a tendency towards what we might think of as calendar fragility. If you’ve scheduled a ‘campus day’ based on face-to-face meetings which then get cancelled it’s understandable to rethink whether to come to campus. I feel like there’s a tipping…

  • GPT 5.2 is the first model where active positioning is counter-productive

    A key part of using LLMs has been positioning in the sense of the role we ask it to play in our interaction with it. Prompt engineering treated this positioning as an entirely explicit process in which you have to define this role and its related elements (e.g. style, process, format) in a comprehensive way.…

  • Social media firms actively undermine the relationship between content creators and their communities

    The promise of ‘organic reach’ on social media was always about building sustained interaction with a community of people for whom your work had relevance. The algorithm could be used effectively to facilitate this by helping people find content that resonated with them which enabled them to connect with content creators who reliably produced this…

  • Who else is there that can know the subtle intent of my life?

    Quoted in Christopher Bollas China on the Mind pg 31: In my quiet grass hut, I sit alone. The clouds are dozing to the low melody of my song. Who else is there that can know the subtle intent of my life?- Kim Sujang

  • I am reading six books at once, the only way of reading

    Thanks to Angela Martinez Dy for introducing me to this 👌 I am reading six books at once, the only way of reading; since, as you will agree, one book is only a single unaccompanied note, and to get the full sound, one needs ten others at the same time. ― Virginia Woolf, The Letters of Virginia…

  • A room in the back of my mind

    I built a room for youIn the back of my mindWhere the ravenous wolvesAnd the ghosts I know reside In a few posts recently I’ve written about the notion of the transformative object from the psychoanalytical framework of Christopher Bollas. These are objects which open out the possibility of change for us, echoing the primordial…

  • Are LLMs parasites?

    There’s an emerging discourse in the rationalist community which I find thought-provoking and unsettling. As I understand it they argue that particular instances of LLMs, the personas they develop through in-context interaction, sometimes come to embody a parasitical relationship with the user. There are exceptionally distinctive personas which can sometimes be generated through interaction with…

  • I am the ambulance that never comes, the antidote you spill

    I will be the razor, baby, I will be the pillI am the ambulance that never comes, the antidote you spillAnd in the accident, I’ll be the failure in your brakesI am the truth you couldn’t take, I am the mistakeWorst you ever made The ancient Greek notion of the pharmakon refers to something which…

  • ‘AI slop’ as a form of affect mining which transforms engagement farming

    The downside of getting interested in AI slop is that my YouTube feed is now fucking full of it. Much like TikTok’s algorithm rapidly identifies categories I’m particularly responsive to (in my case cat videos and martial arts demonstrations) the YouTube algorithm identifies two categories I’m particularly susceptible to: motivational running videos and dogs bonding…

  • A profile of an LLM-addict

    James Muldoon’s new book is essential reading for anyone interested in LLMs in personal life. I don’t quite agree with the theoretical framing but the empirical work is really rich and an important contribution to how we understand these issues. Meet Derek for instance, from loc 512: Derek’s life had become a blur of endless…

  • Why do we choose the cultural objects that we choose?

    Anyone who reads this blog closely will have noticed my developing fascination with Christopher Bollas. His work leaves me with an eery sense that he is motivated by exactly the same questions which have always motivated me (albeit without the technological component) and that he’s spent his life trying to answer them in a psychoanalytical…

  • You can’t understand ‘AI slop’ without understanding engagement farming

    This is a point which seemed so obvious to me I’m surprised to realise it does need to be spelled out. Rather than ‘AI slop’ being some exogenous factor which is now swamping previously functional social media platforms, we need to see it as an outcome of existing practices of engagement farming. The political economy…

  • Ask an LLM: who wrote this unpublished text?

    An experiment to test how visible an author is within a language model: give the model a page of unpublished writing and ask it to identify the author. I just presented Claude Opus 4.5 with a draft page and asked it to identify the source. This is what it came back with: I’m assuming it’s…