The Bullshit Premium
Are we living through a unique and temporary Goldilocks period for AI BS before the whole circus eats itself alive?
I still remember vividly the day at school when I realised our teachers were too overworked to actually give a shit about our assignments. I must have been 14 or 15. It was science class. Our well-respected chemistry teacher had assigned us a big end-of-term project - something to do with atomic weights, something well above our level. I remember receiving back my assignment, neatly assembled in one of those plastic folders with the transparent cover, yellow spine and two little foldable metal clips that went through the holes you would punch in the A4 to hold the stack together. My report was a little chunkier than the others, and as it slapped down on the desk in front of me with a satisfying thud, I was delighted to see a capital ‘A’ in red pen, dutifully encircled and complemented with some comment like “Extremely thorough research - well done”.
It was only when I got home that I realised that half the physical contents of the report was made up of blank sheets of paper. Somehow, I had accidentally left my spare stack of pre-punched sheets in the folder, lending it a wholly baseless heft that had not only gone unnoticed but had also clearly contributed to my top-level grade. Well, that and perhaps liberal use of a still-teacher-proof incoming technology called Encarta on CD-ROM. Perhaps you’re old enough to remember?
I was quite naive as a schoolboy. I’m sure my more astute classmates sussed this out well before I did. For me though, the realisation that teachers were just fallible human beings who would rather watch TV at night instead of grinding through 30 pre-pubescent homework assignments was a rude awakening. Looking back, it was probably key to my considerable academic success in the decade that followed. I’m not, and never was, a cheat. And I worked hard at school and uni, I really did. I simply leaned in and learned how to manipulate the academic system to my advantage: highest possible grades for as little work as possible. In this pursuit I became an absolute expert. Fluff, filler, empty pages and bigger font sizes were the basic building blocks of everything from GCSE coursework to my PhD thesis. From there I graduated to judicious and largely random use of semicolons. What the fuck is a semicolon; and where; does; one; use it? I might have an inkling now, but back then I can assure you that neither me, nor my peers, nor my teachers (apart from the English teacher; don’t push your luck) had even the slightest clue; but they look as impressive; as; hell; coming from a 16-year old. In languages, I learnt that you didn’t have to remember proper conjugations if you stuck a ‘helper verb’ in front the of the main verb and just use the infinitive form. I’d never say “I play football” in French or Russian - instead I’d say “I like to play football”. Simple syntactic shortcut that sounds convincing and saves you learning all the irregulars. Pure fluff. I even worked out that I could ace oral exams by just putting on a dumb-ass French accent and mouthing a Gallic-sounding “eeeeuuughhh” for at least 20 seconds in between each haltering interjection about having a punctured tyre on the D959 to Tours.
All this junk - fluff, filler, empty pages, semicolons, empty verbs, meaningless vocalisations - is bullshit. There’s really no other way to describe it. It’s an optical illusion - real, but meaningless. It wasn’t revolutionary twenty years ago, and it’s not now, even though we’re rediscovering it through the lens of AI. This type of BS has always been a badge of honour or a form of intellectual or professional virtue signalling. More bullshit, until now, has always meant more work, more man-hours, more grind - and that made it actually worth something. Or at least others would perceive it that way. David Graeber’s now classic work “Bullshit Jobs” is full of such examples. By his estimate, a massive chunk of the productive economy was (I say ‘was’ not ‘is’, because his book was pre-ChatGPT so it’s much, much worse now) dedicated to the systematic production of BS.
Bullshit as a desired economic output is kind of hard to fathom, yet it’s absolutely everywhere. And again, this is not new. Junior employees have usually been at the forefront of the bullshit economy. Think of the junior analyst at the investment bank who spends his or her days churning out pages and pages of reports that end up on the desk of another junior analyst on the other side of the country whose job is to filter through those stacks of paper in order to deliver a simple nugget of wisdom to middle management: buy, sell, hold, whatever. The hot potato of bullshit that gets passed from junior employee to junior employee simply served to give them something to do, some task they could contort to show their worth to their employer, to make their middle manager look better, to clock up hours in the office - evenings good, weekends better. Those that embraced the bullshit rose through the ranks.
You’d think professional authors would be above these bullshit filler techniques, but far from it. They are the masters. Single-concept fluff pieces are the norm for books in the ‘Smart Thinking’, ‘Business’ and ‘Self Improvement’ niches. So much so, in fact, that Blinkist have built an incredibly successful business out of systematically de-bullshitting these books so that you, the reader, don’t have to wade through 300-plus pages of fluff to get to the core message. You just listen to a 10-minute summary. This example is essentially analogous to the previous scenario involving the junior employees, just with different actors. In both cases, economic value is ‘created’ in the process of wrapping a useful message in a layer of bullshit. In the case of the analyst, it’s literally his or her salary. In the case of the author, it’s the increased perceived value of a 300-page book versus a 15-page book. Picture the scene - you’re browsing the bookstore at the airport (where else does anyone buy books they will never read?). Are you willing to pay £15 for a pamphlet? Of course you’re not. Like everyone else, you expect a good hunk of wood-pulp for your money. Something that will look good on your shelf. Bullshit creates perceived value. But the ‘bullshit-value-creation-chain’ doesn’t stop there. No, bullshit gives twice. Blinkist creates value by stripping away the bullshit for you, just as the second analyst does for the ‘intelligence’ they present to their manager.
Really, when you look at it like this, is it any wonder that bullshit jobs have been eating the world for the past 30 years and we’re drowning in a sea of slop? You’d think the free market would have dealt with this by now - after all, paying the price of bullshit twice over hardly sounds like the epitome of economic efficiency. A theoretical company that doesn’t cloak their internal knowledge in 27 powerpoint slides, with all the human costs involved in such a process, will presumably outcompete others in their sector by a large margin once scaled up. But that would be to assume that humans can detect bullshit when they see it, which they can’t, at least reliably. Anyway, we’re not ‘rational actors’, as we are all well aware of, so it’s unsurprising that this preference for BS has emerged. In fact I might coin a new cognitive bias: Bullshit Premium. The Bullshit Premium is the difference in perceived or actual value between a ‘clean’ intellectual output, and the same output wrapped in a thinly disguised veil of BS.
In 2025 we find ourselves at an interesting crossroads, BS-wise. There are two opposing forces at play which will ultimately determine which direction this goes and whether bullshit will whither and die, or multiply and thrive. I have hoped, since the early days of ChatGPT and throughout the subsequent iterations, that AI would drive the Bullshit Premium to zero. Think about it like this: we now know to a pretty high degree of certainty now that any long, thorough and well-written piece of work, like a research report, executive summary, or strategic plan, will have been researched and written by AI, with an absolute minimum of human input. Oh pipe down for god’s sake. Don’t try to pretend you’re any different. Everyone knows. Which raises the question: if everyone knows, why the hell would a bullshit-cloaked piece of work hold any perceived value over and above the original idea? It shouldn’t. It’s absurd that it still would. Nobody, literally nobody, is impressed with your 20-page report. So just stop. Give us the abstract. Give us the original creative spark. Hell, to paraphrase Clayton Ramsey - just give us the prompt. Bullshit no longer requires human toil. Bullshit created by a Harvard graduate doesn’t reflect positively on your company any more than 3-ply toilet paper in the head office toilets. Bullshit is now a total and utter commodity, costing fractions of a cent per kilo. It should just disappear naturally, like manure.
The counter-argument to this is that everyone over the age of 40 still thinks bullshit looks and sounds impressive (or at least assumes that everyone over the age of 50 thinks that way). They think they might just be able to trick their boss or teacher into thinking they did some actual ‘work’ if they add ‘make it sound original’ to their prompt. The result is looking like a decade-long LLM bender and a yet another example of the prevalence of a digital commodity exploding as its intrinsic value goes to zero. Like porn, bullshit is now free, so we gorge on it. Eventually we will have re-bullshitters attached to our mouths and de-bullshitters attached to our ears such that you will never even be aware of the existence of the bullshit in the first place - it will be fired across the ether between two devices manufactured by OpenAI, Google or Apple - a BS-encoded bytestream costing a bargain 13.99/month and a global 10 Petawatts of electricity per year. That’s basically what’s happening now when you use ChatGPT to create a document that I then have to feed back in for summarisation. It’s a pain in the arse and costs us both $20 a month.
I think it’s worth repeating: I know you didn’t write that yourself. I am NOT your grandma. I am NOT impressed.
Bottom line: wrapping some genuinely useful idea in a shell of unwanted BS is not new. It was invented donkeys’ years ago and gradually perfected by every passing generation of writers, academics, junior researchers and students, including yours truly. AI didn’t invent this shit. But it sure as hell has perfected it. What, then, lies in store for us humans if even bullshit jobs are no longer on offer? My current feeling is that the near-term future will be dominated by those that can pull off the AI bullshit trick most convincingly. I think there’s still enough of a Bullshit Premium for the grifters to go after because there are still enough people who will be tricked into (or straight up don’t care about) reading books written by AI, watching videos scripted and performed by AI avatars, signing up for $5/month software platforms that scratch some niggling little digital itch they never knew they had, or buying some product being pushed by an influencer that doesn’t really exist.
Perhaps we’re living through a unique and temporary Goldilocks period for AI BS before the whole circus eats itself alive.


Actually makes me feel bullish on the AI economy for the first time in a while.
If bullshit jobs are such a massive part of the economy for so long, maybe there is a $20 trillion market “prime for disruption” aka a real revenue stream for OpenAI and pals. And then charging us double for the de-bullshitter.
This is a genius idea to make money. Not so great for actual creative thought or evolution of culture, but maybe this means the stock market won’t crash before Easter.