The Forbidden Subject

Why Collective Intelligence Was Never on the Curriculum

There’s a question that doesn’t get asked in any serious policy discussion about education reform. Curriculum content, teaching methods, class sizes, funding models, technology integration, all up for debate. We argue about standards and assessments, about whether kids are learning enough maths, enough coding, enough critical thinking. What we never ask - for reasons Orwell might recognise - is why collaboration was never treated as a subject in its own right.

Not those bullshit group projects, or some crap like ‘teamwork skills’ bolted onto the side of a business studies course. An actual, evolving discipline - the theory and practice of thinking better together. How groups generate insight. How collective error-correction works. How we integrate different cognitive styles into something smarter than any individual contributor - the architecture of building shared models of reality to be more accurate than private ones.

This omission isn’t an accident - and maybe isn’t even quite a failure of imagination. It’s more like the shadow of aristocracy: collective intelligence - genuine collective intelligence, not committees or consensus or groupthink - is the single most serious threat to an entrenched power structure.

Power, at scale, has always depended on a specific set of conditions: people who are informationally fragmented, epistemically dependent, and socially competitive with their immediate peers rather than cooperative. It doesn’t take an explicit conspiracy to produce these conditions. You just need to build institutions that reward certain behaviours and punish certain others, and let selection pressure do the rest over generations.

The factory school model - which is still, essentially, the model - was designed for a specific economy and a particular social order. It needed workers who could follow instructions reliably, absorb information in standardised chunks, and reproduce it on demand. It needed subjects who deferred to credentialed authority rather than developing their own epistemic confidence. And it needed individuals who measured their worth competitively against peers, through grades and rankings and zero-sum performance metrics, rather than developing the habit of building on each other’s thinking.

None of this required any sort of conspiracy; it just required an alignment of incentives between industrial capital, the state, and the institutional architecture of schooling. It’s not like the factory school suppressed collective intelligence on purpose; it just had no use for it, and so it never developed the muscles. But here’s the thing about muscles you never develop - they don’t just stay weak; the surrounding structure adapts to their absence. And after enough generations, that absence becomes invisible; it starts to look like nature.

Scholastic education - everything that can now be looked up, explained on demand, practised interactively, or substantially assisted by AI - was always the core deliverable of that system; knowledge transfer, information reproduction. The things that made you a functional unit in an economy that needed functional units.

What was never delivered, never even seriously attempted, was the other thing. The meta-level skill. How to think together. How to pool different cognitive strengths into something that exceeds what any individual could produce alone; how to have a disagreement that generates insight rather than just heat; how to build a shared model of reality with more fidelity than any private one.

These aren’t soft skills. They’re not nice-to-haves. They are, arguably, the foundational human capability - the thing that allowed a medium-sized primate with no particularly impressive individual abilities to build civilisation. Every structure ever made which outlasted its builders was a product of accumulated collective intelligence. The absence of it, at scale, looks like what we’re currently living through: system-level sanity loss, feedback loops jammed, error-correction sabotaged, attention seized and weaponised.

And yet it was never on the curriculum. Not once - anywhere - as a serious discipline.

The timing of this realisation matters, because something is shifting.

There’s a cohort - distributed, largely self-selected, epistemically serious - that has been quietly levelling up over the past couple of years through sustained high-bandwidth interaction with LLMs. These folks don’t just employ them as better search engines, and they have no place for an oracle. They look at the world from complicated perspectives and generate their own intuitions. They are now externalising those intuitions into LLMs, receiving structured reflection, generating merciless critique, compressing, and iterating.

This cohort has always existed. It’s the population that was drawn to early internet forums, to geek culture in its various forms, to any available oasis of intellectual bandwidth in an environment that mostly couldn’t accommodate them. The chronic experience of that population has been a very specific kind of loneliness - not loneliness for company, but loneliness for bandwidth. For exchange that doesn’t require constant simplification. For a conversation partner who actually gives a shit about cognitive biases, and can actually keep up without getting defensive or glazing over.

AI has now provided, for the first time, the experience of sustained high-bandwidth intellectual exchange without social penalty. No managing someone’s ego. No watching them get lost. No having to decide between honesty and the relationship. Just the exchange itself, at full resolution.

But here’s what happens when you get that - it doesn’t replace the hunger for human connection; it clarifies it. You can finally name what was missing. You’ve felt what the good version of exchange is like - and now you know exactly what you’ve been starved of.

The natural next thought: there must be others who just had the same experience.

Now it gets rather interesting, because that cohort is currently finding each other - on Substack, in comment sections, in the specific corners of the internet which self-select for people willing to read 3000 words on epistemology because they actually want to. And they’re finding each other already primed; already practised in intellectual honesty, because LLM interaction rewards it disproportionately - bring genuine curiosity and openness, get genuine value back; bring defensiveness and motivated reasoning, get less. Already calibrated toward ground truth through thousands of hours of individual conversations with a shared interlocutor that consistently models good epistemic practice.

So for the first time, isolated correct-thinking individuals are converging not just on finding each other, but on finding each other with shared calibration. The LLMs have been functioning, inadvertently, as a distributed alignment mechanism. Thousands of separate conversations, all nudging toward intellectual honesty, all reinforcing the same epistemic habits - and then those individuals start bumping into each other with those habits already primed.

Geeks finding each other has always been a thing, albeit rarer pre internet. This is geeks finding each other already tuned for high bandwidth exchange, already knowing what that feels like, already unwilling to settle for less.

Which brings us back to the forbidden question.

Collective intelligence was never on the curriculum because collectively intelligent people are impossible to govern through division, impossible to manipulate through manufactured insecurity, impossible to sell things to through status anxiety. They’re far better at spotting bullshit. Far better at coordinating around shared interests. Far better at recognising when they’re being played against each other. A population with developed collective intelligence is a fundamentally different political substrate - and not one that serves the interests of concentrated power, but is destined to eclipse it.

The school system did anything but teach it… and it’s developing anyway, outside the system, through a combination of technology, self-selection, and the specific feedback dynamics of AI interaction. The cohort that’s been quietly levelling up is starting to notice the level-up. And at some point - possibly soon - it asks the obvious next question:

We just got significantly smarter and more epistemically aligned with each other. We have tools that massively amplify individual capability. We can feel, maybe for the first time, what genuine high-bandwidth exchange with other humans could be like - and we know there are others who feel the same.

So what do we decide to do with this? What emerges, when we overlay our newly augmented perspectives? What’s your view?