humanworks.

Tag: chatgpt

  • The Cost Of Counting – Losing The Apple Watch And Trying To Not Lose My Mind In AI Cost Calculations

    The Cost Of Counting – Losing The Apple Watch And Trying To Not Lose My Mind In AI Cost Calculations

    Ten days ago I did something that, for a person who describes himself as a metrics driven autistic with a chronic pain condition and a relationship with sleep that could generously be described as “complicated,” might have seemed genuinely unfathomable.

    I took my Apple Watch off.

    Not just to jump in the shower. Not because the battery died. Not because the strap was chafing, although if you’ve ever spent three days staring at a sleep score whilst simultaneously developing a mild repetitive strain injury from anxiously rotating your wrist to check your sleep score, you’ll understand that chafing is the least of your problems.

    I took it off and put it on my bedside table. Then I thought about whether to put it back on. Then I didn’t.

    I haven’t worn it since.

    The reason was a conversation with a good friend – one of the people who knows me well enough to know about the fibromyalgia, about the sleep apnoea diagnosis that arrived in the same window as a global pandemic as though the universe decided that if it was going to do bad timing it might as well commit to the bit – and we were talking about my sleep score and how I could improve it.

    My sleep score was a conversation. I had plumbed new depths.

    If you’ve used any consumer health wearable in the last five years, you’ll know exactly the flavour of psychological torture I’m describing.

    Every morning, before you’ve had coffee, before your nervous system has fully negotiated the terms of another day, you’re presented with a number that tells you how well your body performed its most fundamental biological process whilst you were unconscious and therefore unable to do anything about it. This number is presented as the truth in the same way as a judge might deliver a sentence – and the sentence was broadly “do better, Matthew”.

    For me specifically – and I’ll acknowledge upfront that my relationship with numbers is somewhat more intimate than most people’s, which is both a superpower and a source of entirely self-generated misery that I’ve had to spend considerable therapeutic energy untangling – the sleep score had become its own recursive problem.

    I was tracking my sleep quality in order to manage my sleep quality, which was being negatively affected by the anxiety generated by tracking my sleep quality. Getting told you’ve had a bad night sets you up to believe you have had a bad night, even if you didn’t feel it to start with. The problem there is that I’m relying on a machine more than my own thought process – something I’ll come to later in this article.

    The insight that finally shifted things was less about some massive realisation. Instead, it was embarrassingly simple. I was explaining to my friend why I was miserable about my sleep score, and I heard myself saying, out loud, that the score was making me more anxious about sleep, which was making the sleep worse, which was making the score lower, which was making me more anxious. I’d built myself a perfect closed loop of quantified suffering. A feedback spiral with excellent data integrity.

    She was giving me advice on how to improve my score, whereas I felt like I wanted to just… well see what happens without a number to chide me further.

    I took the watch off straight after that chat.

    the inverted machine, or: whilst i was trying to stop being a machine, a machine was trying to become me

    Now to talk about what is related, but also inverted – the difference between a machine tracking me, and me tracking a machine as the business has invested heavily in Codex and the broader OpenAI partnership.

    In the same fortnight I divested from my personal biometric surveillance apparatus, I found myself spending a rather significant amount of time – intellectually, financially, and in terms of the conversations I was having with my colleagues – on the question of machines. Specifically, on what it costs to operate them. Even more specifically, on why that cost is considerably harder to calculate than the people selling you the machines would like you to know.

    The contrast, when I finally noticed it after our weekly retro at work on Friday, was almost too neat for a person who enjoys a good structural irony.

    I had been the human attempting to quantify myself like a machine. Importing the logic of the dashboard, the metric, the score, onto biological systems that were not designed to be measured that way and were communicating their objection through the medium of increasingly poor sleep scores and the particular brand of 3am anxiety that feels like your nervous system has hired a crisis communications team.

    Simultaneously, I had been watching – and paying for – a machine that is very keen to have you believe it’s something closer to human than it actually is.

    Not because it’s malicious, and not because the people building it are deliberately deceptive (though we can revisit that particular conversation later), but because the cultural framing around it has collectively decided to conflate output with intent in a way that we really should know better than by now.

    We have all, at some point, been the child saying sorry for the thing we weren’t actually sorry for. We understood instinctively that producing the appropriate social output does not necessarily constitute the internal state the output implies.

    The AI is not doing that consciously. Which is, in some ways, worse.

    We are anthropomorphising a stochastic process because the outputs are fluent and occasionally uncanny, which is exactly as robust as deciding your washing machine has a personality because it sometimes makes an unusual sound during the spin cycle. The groan isn’t the washing machine expressing ennui – it’s a noise.

    Anyway. Machines. Costs. On to, the Friday conversation and why you can’t calculate them as simply as you’d like.

    tokens: not one thing, and definitely not as cheap as advertised

    Last Friday I was sitting in a catch-up with Steve, who runs the business I now co-lead on the technical advisory side, and he asked a question that is increasingly being asked by senior leaders across every organisation currently in the process of discovering that deploying AI at scale costs rather more than the pilot suggested.

    “How do you pre-calculate token usage to avoid spending excessively?”.

    OK, technically it was “does anyone know how to present the relationship between price, OpenAI credits, and tokens” but the underlying question was the same.

    So how do you know what you’re going to spend? Is it possible to assess ahead of time?

    It’s a fair question. It’s also a question that, once you start answering it properly, reveals that whilst you might be able to align the logic of “one credit equals 150 tokens” the truth is that tokens aren’t able to be pre-calculated when you use a thinking model.

    One of my colleagues who was also on the cal, made a reasonable attempt at an explanation. It wasn’t quite right. This is not a criticism of them – it’s genuinely not an obvious thing, and the way the industry presents it doesn’t help, because the industry’s commercial interests are served by keeping certain aspects of the cost structure somewhat opaque. We’ll come back to that. On that side, that’s a point my colleague was bang on about – it’s less about value assessment, and more about “just one more token” which is the architecture of illegal product distribution mechanisms.

    Or, to be put it bluntly, it has the same psychological profile as that of a drug dealer – don’t worry about the cost, just enjoy it.

    Anyway, getting back to the mundanity, let me explain what tokens actually are, because the word is being used in about four different ways simultaneously and this is, I would argue, not entirely a coincidence.

    A token is, in the simplest possible terms, a unit of text. It’s roughly – and I want to stress roughly, because precision here is itself part of the problem – about three-quarters of a word in English, though this varies considerably depending on what you’re asking the model to process. Code is tokenised differently to prose. Non-Latin scripts tokenise differently again. If you’ve ever wondered why querying AI in certain languages costs disproportionately more, this is part of the answer.

    Now. Input tokens and output tokens. These are the ones most people have a vague handle on. You send the model some text (input tokens, charged at one rate). The model sends back some text (output tokens, charged at a different – typically higher – rate). Simple enough. Calculable in advance if you know your prompt length and can estimate your expected output. Not trivial to predict with precision, but manageable.

    So you can work it all out in advance? Hold your horses a little.

    This is where I tell you about thinking tokens.

    Ultimately, this is where the clean ledger gets complicated, and where the consultancy parallel of my own career is relevant.

    internal monologue is expensive: the thinking model problem

    When you ask a thinking model to work through a complex problem, it doesn’t simply process your input and generate an output. It does something considerably more interesting, and considerably more expensive.

    It talks to itself first.

    The technical description is “chain-of-thought reasoning” or “internal scratchpad” depending on which company’s documentation you’re reading, but the phenomenological description is closer to what I’d call a person’s internal monologue. The model generates reasoning steps that it doesn’t necessarily surface to you in the final output, works through competing hypotheses, revises its own approach mid-thought, and arrives at an answer that was produced by a process you can’t fully observe and which – critically – consumes tokens you are paying for whether or not they appear in what you receive.

    I explain this to people using a consulting analogy, because I work for a consulting firm and it seems appropriate to use shared vocabulary.

    When you retain a consultant, you’re paying for their time. That time includes the hours they spend in meetings with you, the deliverables they produce, and the thinking they do that never appears on a slide.

    The good ones – the ones who arrive at answers that seem almost intuitive in retrospect – are often the ones doing the most invisible work. They’re running problems in the background whilst apparently doing something else, noticing patterns that don’t fit, interrogating their own assumptions before they become your advice.

    That internal processing is billable, even when it’s unconscious. You don’t see it. You see the output. But the expertise that shaped the output was built in the invisible thinking, and you’re paying for that expertise even when it’s not legible.

    This is why my day rate is what it is – you’re not paying for a day of my time, you’re paying for the experience I have that means you only have to spend a day doing something rather than months struggling through treacle making the mistakes I have already made years ago.

    So when it comes to using a thinking model, you get to talk to something that can talk to itself about experiences – although this time, the experience is their data set as opposed to the lived experiences of consultants. Only time will tell if these two things are suitable delineated to show the value of consulting over time.

    Thinking models work the same way as consultants, except you’re not paying for the expertise directly – instead you’re paying for the actual compute consumed by each step of the internal reasoning process. The challenge is the number of steps isn’t fixed, isn’t easily predictable, and isn’t consistently exposed by the platforms you’re deploying through.

    When someone speaks to me, they don’t get to see my internal reasoning – which makes me very much like an AI model, except I do at least have the decency to tell you the day rate first.

    Most people want to know that “you can ask 100 questions a week” because that fits well on a financial ledger. The reality is it isn’t that clear. The deeper problem is that not all usage is created equal – just as we’ve seen people 10x their productivity, and we’ve seen others make images about Cthulu.

    Anyway, getting back to the calculations… even if we broke down the input tokens – which you can estimate with reasonable accuracy – the thinking tokens and output tokens are functions of what the model decides to do with your prompt, which is itself a function of prompt complexity, temperature settings, and model behaviour that most cloud-based deployments don’t give you clean visibility into.

    Temperature, for those unfamiliar, is roughly the dial between “deterministic and predictable” and “creative and variable” – and higher temperature means higher variance in output length, which means higher variance in cost.

    I don’t generally have time to sit down and talk about the finer points of AI model parameters in chats because most people aren’t actually interested in the mechanics. However, I’m getting into it now, because this is a blog post rather than a two-minute summary and I have considerably more latitude.

    There are ways to spend less than the meter implies — getting the model to hand you a deterministic script that runs without calling it again, rather than paying for a fresh invocation every single time.

    The commercial reasons the platforms would rather you didn’t dwell on that option are a piece in their own right, and a later one. For now it’s enough to sit with the smaller, stranger fact: the cost structure resists clean calculation, and that resistance is not an accident.

    Or, in simple terms, think before you prompt.

    the unified problem: when does measurement serve you, and when does it consume you

    Here is the thing I’ve been working towards.

    More data creates more pressure to measure what we are doing. I’m already seeing it with clients. Everyone wants clear pricing and clear cost control the same way they wanted to do this when the business moved to the cloud and realised 100 people doing the same thing 100 times over was not cheap. However, AWS at least had the decency to tell you what an hour of compute cost.

    So this is not a new observation, but it’s worth sitting with in the context of both of the things I’ve been describing, because the vector of the problem is different even as the structure is identical.

    With the Apple Watch, I was the thing being measured. The data proliferation was about me – my sleep cycles, my heart rate variability, my blood oxygen levels, the number of times I apparently shifted position between 2am and 4am. All of that data was real. None of it was particularly actionable beyond “sleep better,” which I was already motivated to do and which the data was, empirically, making harder rather than easier. The measurement was consuming the thing it was supposed to serve.

    With AI spend, the organisation is the thing being measured – or rather, the organisation would like to do the measuring but can’t, cleanly, because the cost structure actively resists clean measurement.

    Steve’s question was the right question. The honest answer is that you can get close with careful prompt engineering, with moving deterministic workloads into scripts, with understanding which use cases benefit from thinking models, and which are overpaying for probabilistic variance they don’t need.

    However you cannot fully pre-calculate AI spend with the transparency a finance function would like, because the architecture doesn’t expose the parameters cleanly, and the commercial interests of the people building the architecture are not aligned with exposing them.

    These are different manifestations of the same dysfunction. In one case, the data is too present and too actionable-feeling, creating anxiety loops that degrade the system being measured. In the other, the data is deliberately obscured, creating cost overruns in systems that were sold on efficiency.

    There’s a third face of it, and it’s the one that closes the loop back to where I started.

    A thinking model doesn’t always know when to stop thinking. Past a certain point, more recursive self-questioning buys you more hedged, more verbose, occasionally more confused output rather than a better answer – and the boundary between “this benefits from extended reasoning” and “this is disappearing up its own chain-of-thought” is not marked on any dashboard, while the meter runs either way. I say this as someone who has been reliably informed by multiple therapists that I have precisely the same problem. The model over-measuring its own reasoning is me over-measuring my own sleep: neither of us knows when enough is enough, and the not-knowing is the expensive part.

    (Cynics may well say “this sounds a lot like you when you’re talking philosophy in a tech context” and I’d probably say cynics are right. It’s one of my many idiosyncrasies.)

    Anyway, sanity, in every case, is the same discipline: knowing which data is load-bearing, which is noise, and what the cost of measuring versus not measuring actually is.

    I know what my sleep quality is like without a number. I know when I’ve slept badly because I wake up in pain and my cognition operates at the level of a drowsy golden retriever. The number added precision without adding insight, and traded the precision for my equanimity at 7am.

    I know what good AI deployment looks like without a per-token ledger. A thinking model writing compliance documentation is justified spend. A thinking model answering “what’s the capital of France” is a contribution to OpenAI’s quarterly results that adds nothing to mine. In that case, there’s nothing to “think” about – some things just are, whereas other things need thinking and their associated inference that these models bring us for a subscription price.

    The watch taught me this and cost me nothing except the embarrassment of having taken this long to notice – and the £800 or so for the Apple Watch Ultra.

    The AI cost structures, if they teach the same lesson, will cost rather more to the companies that adopt them before the penny drops.

    Some things, it turns out, are better measured with experience than with instruments. The trick is knowing which is which before you’ve spent three months of compute budget and a year of 3am anxiety finding out the hard way.

    And here, finally, is the symmetry I couldn’t see while I was living inside it.

    I was a human trying to run myself like a machine – importing the dashboard, mistaking the score for the sleep.

    In the same fortnight, I was paying for a machine running the trick in reverse: a stochastic process very keen to have you believe there’s a someone in there, fluent enough that we conflate the output with an intent it doesn’t possess.

    Two lies from opposite ends of the same mirror – mine, that a number could tell me how rested I was; the machine’s, that a confident answer is the same as a considered one, and that its cost, like its interior, is simpler than it looks.

    I stopped letting one machine measure me. The other one is always measuring me back – fluent, confident, and believing I won’t notice. When it comes to using AI effectively, noticing is the whole of the work.

    Next week: what we’re actually doing to push past the current limits of what thinking models can do, why those limits are more interesting than the hype suggests, and why “just use a bigger model” is the AI equivalent of “just try harder” – technically true, immediately useless, and beloved of people who haven’t looked at the cost structure recently.

    Series — The Option Problem

    • Part 1: The Cost of Counting
    • Part 2: The Paralysis Paradox — coming soon
    • Part 3: The Superposition Problem — coming soon
    • Part 4: The Wrong Objective — coming soon

  • You aren’t reading enough, and you definitely aren’t thinking enough – so read and think about this

    I’m reading Henry Fairlie this weekend. Bite the Hand That Feeds You – collected essays from one of the sharpest provocateurs the English language produced, and a man whose photograph on the cover alone – cigarette in hand, glasses slightly askew, typewriter lurking in the foreground like an accomplice – communicates something about the relationship between a writer and their craft that no amount of productivity guru content has ever come close to replicating.

    (The typewriter is doing real work in that image. It isn’t decorative. It is the instrument through which the provocations were forged, and there is something quietly honest about having it visible – no pretence that the words simply materialised from some frictionless creative ether. They were hammered out. Key by key. Which is, when you think about it, rather the point of what follows.)

    Those of you who know my influences will know that Christopher Hitchens occupies a significant position in how I approach both writing and argument. Not because Hitch was provocative – though he demonstrably was – but because his provocation was deployed with genuine intellectual scaffolding beneath it, which is a distinction that matters enormously and that most people confuse with volume. You don’t awaken someone from the torpor of collective slumber with a gentle suggestion. You use a bucket of cold water. The trick – and it is a trick, albeit requiring genuine craft – is ensuring the bucket contains substance rather than merely noise.

    Fairlie understood this. Hitch understood this. Whilst in an age where we have outsourced the generation of text to systems that are, by any honest assessment, genuinely impressive at producing words whilst being fundamentally incapable of the thing that makes words matter, understanding this distinction has become rather more urgent than it was when Fairlie was bashing away at his typewriter.

    the agent provocateur’s actual job, or why being uncomfortable is the point

    Fairlie’s polemics were, I suspect, constructed partly for effect – closer in spirit to the work of an edgy comedian than to some earnest manifesto designed to reshape civilisation overnight. There is nothing wrong with that assessment. In fact, there is something deeply undervalued about it, because it misunderstands what the effect actually is.

    Here’s the thing.

    Understanding how to construct an argument – not merely to have an opinion, which is approximately as difficult as breathing and roughly as intellectually demanding – but to deploy that opinion with knowledge, precision, and persuasive architecture that forces the reader to genuinely engage rather than simply scroll past – is one of the foundational skills of anyone who wants to make a real impact on anything beyond their immediate surroundings.

    I learned this in amateur debating societies, where the single most valuable lesson was not how to win an argument but how to understand the opposing position well enough to articulate its strongest case and then use that knowledge to dismantle them.

    Those who cannot do this aren’t debating. They’re performing. The distinction matters because performance can be detected, dismissed, and scrolled past in approximately 0.3 seconds. Genuine argument – the kind that actually lands – requires the reader to do cognitive work. It requires friction. By comparison, spouting rhetoric – that pervasive performance that many think sits as some actual substitute for argument rather than the piss poor presentation of idiocy – is not debating at all.

    Fairlie understood this instinctively. His essays don’t simply assert positions – they construct them with enough rigour and enough provocation that the reader finds themselves genuinely wrestling with the ideas rather than passively absorbing them. The discomfort is not a bug. It is, in the most literal sense, the mechanism by which thinking actually occurs.

    (And yes, I recognise the recursion here – I am arguing, via essay, about why essays matter, whilst simultaneously doing the thing I’m describing. My therapist, Becky, would note this with a raised eyebrow and the observation that “Matt is doing the recursive analysis thing again.” She would be correct. The recursion never stops. Welcome.)

    the cognitive friction complex™ (or a lack thereof)

    We live in a moment of extraordinary and largely unexamined paradox regarding information and capability. We have, quite literally, more collective knowledge accessible through our fingertips than at any previous point in human history. Simultaneously – and this is the part that deserves rather more attention than it currently receives – the tools now available to generate text on our behalf have created an environment where the process of engaging with ideas is increasingly being outsourced to systems that, whilst impressive in throughput, cannot replicate the cognitive friction that actually changes how you think.

    This matters more than most people appreciate. Considerably more.

    The ability to cultivate not merely awareness of information but the capacity to use it effectively – to construct arguments, to identify the weak points in positions that appeal to us, to hold genuinely opposing views in tension without immediately dismissing them as wrong because they’re uncomfortable – is a skill that degrades with disuse. It is, in this sense, rather like physical fitness. Nobody loses the capacity to run by deciding not to run once. The degradation is gradual, imperceptible, and by the time you notice it, you’ve lost ground you didn’t know you were standing on.

    Erudition – whether formally acquired or built through the kind of autodidactic discipline that involves actually sitting with difficult texts until they yield rather than asking an LLM to summarise them – isn’t a luxury. It’s infrastructure. Cognitive infrastructure, specifically, and infrastructure that societies require to function at anything beyond the level of collective reflex.

    Now, here’s where it gets interesting. And by “interesting” I mean “slightly existentially destabilising if you follow the thread far enough, which I obviously intend to do.”

    (Stay with me.)

    what an LLM actually does, and what it doesn’t

    An LLM – a large language model, for those who have somehow avoided the last three years of breathless discourse on the subject – is, at its core, an extraordinarily sophisticated pattern-matching system. It has consumed vast quantities of human-generated text and learned to predict, with remarkable accuracy, what sequence of tokens is most likely to follow any given input.

    This is genuinely impressive. I say this without irony or false modesty on behalf of the technology. The statistical inference involved is staggering, and the outputs are frequently useful, occasionally insightful, and – in the right hands – genuinely productive.

    Here is what an LLM does not do.

    It does not think. Not in the sense that Fairlie thought when constructing his provocations, or that Hitchens thought when dismantling an opponent’s position with surgical precision. It does not experience the cognitive friction of encountering an idea that genuinely challenges its existing framework – because it has no existing framework in the sense that you or I possess one. It has statistical weights. These are categorically different things, in much the same way that a photograph of a fire is categorically different from an actual fire, despite being visually recognisable as one.

    (The photograph will not warm your hands. The LLM will not change your mind. Both will give you the impression of the thing whilst being, in some fundamental sense, the total absence of it.)

    What an LLM produces when asked to write an essay is therefore not an essay in the sense that Fairlie wrote essays, nor the way that Hitch did, or how I do.

    Instead, it is a statistically probable approximation of what an essay looks like – the textual equivalent of a very convincing forgery. Smooth, competent, occasionally even elegant. Entirely devoid of the thing that made the original worth reading in the first place. It’s technologically driven sophistry with the depth of a puddle.

    The thing being: a consciousness grappling with something it found genuinely difficult, and producing language as a byproduct of that grappling.

    which brings us back to the question of what reading actually does

    Here is an uncomfortable observation that I have been turning over for some time, and which Fairlie’s essays have crystallised rather neatly.

    When you read a genuinely provocative essay – one constructed by a mind that was actually wrestling with the ideas it presents – something happens in your own consciousness that is categorically different from what happens when you read competent but friction-free text. Your assumptions get disturbed. Your pattern-matching gets interrupted. You are forced, briefly but genuinely, to consider a perspective you hadn’t previously entertained, and the cognitive effort of doing so leaves a trace.

    This is not metaphor. This is, in the most literal neurological sense, how minds change. Not through passive absorption of information – which is what scrolling, summarising, and LLM-assisted reading largely provides – but through active engagement with ideas that resist easy consumption.

    Sometimes people can consider my non-dualistic thinking to be the rough equivalent of getting splinters in my arse as I sit on the fence. In reality, it’s nothing like that – it’s just having an openness to be able to let in the message of things that are being said, not only because it may make your ego feel vulnerable as new data arises, but specifically because we should seek to challenge what we think with the tools of finding what is right.

    In short, to learn you have to accept the reality that you may be wrong and move on from that rather than entrenching yourself in a position. It’s deeply uncomfortable, stirs up emotion, and is prone to make you wonder what’s going on – arguably the opposite of what our increasingly intellectually soporific state offers as the easy option.

    Sometimes you need a wake up call. Fairlie’s essays resist easy consumption with the subtlety of a sledgehammer to the temple. As essays, they are deliberately constructed to create an impact. The provocation isn’t decoration – it’s the mechanism of delivery. The discomfort is the point of entry telling you to wake the fuck up.

    (Which raises a question that I find genuinely fascinating, and which I’ll pose here before I disappear down the rabbit hole it opens – which, knowing my brain, I absolutely will: if the value of an essay lies not in the information it contains but in the cognitive friction it generates in the reader, then what happens to that value when the reader outsources the reading to a system that experiences no friction whatsoever? The information survives. The transformation does not. And it is the transformation that was ever the point. In short, we end up with well written but pointless AI photocopies of thinking whilst thinking goes the way of the dodo)

    the attention span question, handled honestly for once

    The conventional narrative about attention spans runs something like this: they’re shrinking, long-form content is dying, the future belongs to thirty-second video clips and algorithmically optimised dopamine delivery systems designed by people whose own attention spans are, presumably, slightly longer than the products they’re creating.

    This narrative is partially true and almost entirely beside the point.

    Yes, the average attention span appears to be contracting – though one might reasonably question whether it ever existed in the unified form we nostalgically imagine, or whether we’ve simply become more honest about the distribution. The person genuinely engaged with something they care about will still read five thousand words. They always have. What’s changed isn’t human cognitive capacity but the competition for the first thirty seconds of attention before someone decides whether a piece of writing deserves the effort of genuine engagement.

    (If you like my work, you’ll take the time to appreciate it. Others will bounce at just seeing the word count and that’s OK too – although I’d argue that they need to find topics that they find sufficiently interesting to keep their own attention spans healthy without implying my work is going to be for everybody. By design is explicitly isn’t, and is designed to create discomfort in much the way as my mate Tom’s gut reaction is to mushrooms albeit with less toilet based carnage)

    The real question – and this is the one that actually matters – isn’t whether long-form writing will survive as a format. It’s whether the capacity to engage with it will survive in sufficient numbers to maintain the intellectual commons that civilisations actually require to function.

    This isn’t abstract philosophising. This is a structural question about the cognitive infrastructure of societies.

    Fairlie’s essays represent exactly the kind of material that either sharpens one’s capacity or reveals the absence. There is no middle ground with genuinely provocative writing. You either engage with the argument and find yourself thinking differently afterwards – which is to say, you find yourself changed, however slightly – or you bounce off it immediately because the cognitive infrastructure required to absorb the friction simply isn’t there.

    The uncomfortable bit follows.

    The capacity to absorb that friction – to sit with an argument that challenges you, to resist the impulse to dismiss it because it’s disagreeable, to actually do the work of understanding why an intelligent person might hold a position you find uncomfortable – is itself a skill. A skill that requires practice. A skill that atrophies without it.

    Essays are one of the primary instruments through which that practice occurs.

    the uncomfortable implication, or what fairlie actually teaches you in 2026

    Here’s what reading Fairlie in 2026 actually teaches you, stripped of nostalgia for a different era of political discourse and stripped, equally, of any romanticised notion that things were better when writers bashed away at typewriters whilst smoking in black and white photographs.

    (I will, unashamedly, claim my preference for one of Hitch’s favourite drinks – Johnnie Walker’s Amber Restorative – but acknowledge that as one of my role as a Gen X/millenial whereas many young people will see such a tipple as equivalent to chain smoking Marlboro in the 1970s)

    Getting back to the study of essays, it teaches you that the ability to write well about something – to construct prose that forces genuine intellectual engagement rather than merely confirming what the reader already believes – is vanishingly rare, increasingly undervalued, and arguably more important now than at any previous point in history.

    Not because we lack information. We are drowning in information. It’s literally everywhere and injected into your eyeballs at ever increasing speeds.

    Not because we lack the tools to generate competent text. We have more of those than ever.

    Because we are, as a civilisation, systematically undermining the very cognitive capacity that makes information meaningful – the capacity to be changed by it. This in particular is the Achilles heel of modern LLMs – they are architectural designed to kiss your arse so hard it may leave a mark. Essays, by contrast, tend to leave a mark of intellectual whiplash when they are deployed correctly.

    Instead, we have unprecedented tools for generating text. We have, comparatively speaking, a dwindling investment in developing the human capacity to think through text rather than merely consume it. The essays of someone like Fairlie represent the product of a mind that did the latter extensively and the former with genuine craft – a mind that understood, whether consciously or instinctively, that the value of writing lies not in what it tells you but in what it does to you.

    (And here, if I’m being honest – which I am, because this is a version of an essay I’m putting on my website and not LinkedIn given the whole point of this platform is that I don’t have to pretend otherwise – I should note that writing this essay has done precisely that to me. It has forced me to articulate something I’d been circling for months without quite landing on. The cognitive friction works in both directions. The writer is changed by the act of writing, and the reader is changed by the act of reading, and neither transformation is possible without genuine resistance. Without difficulty. Without the uncomfortable sensation of ideas that don’t slide smoothly into place.)

    If you value that kind of intellectual friction – the productive discomfort of encountering an argument that genuinely challenges your assumptions – Bite the Hand That Feeds You is well worth your weekend. The political context is historical, certainly. The underlying skill on display – how to make someone actually think – is timeless. Although one might argue that the desire to challenge the political status quo is needed now more than ever.

    That skill of writing is worth studying. Worth practising. Worth protecting from the comfortable assumption that competent text generation is the same thing as meaningful writing. It isn’t – and I’ll strongly argue it never will be.

    LLM sophistry is not an essay and it isn’t designed to provoke. The difference between writing content and actually changing opinions through discomfort might be one of the more important distinctions of the next decade.

    The world moves forward. How we choose to respond is in our hands.

    Do me one favour, ideally before we collectively forget how to think.

    Read the fucking book.