The Future Belongs To The Weird

出る杭は打たれる — the stake that sticks out gets hammered down.

— Japanese proverb

Every culture has a version of this sentence. The Dutch say doe maar gewoon, dan doe je al gek genoeg — just act normal, that's crazy enough already. The Russians have a whole genre of jokes about the fate of the one who raised his hand. Australians call it tall poppy syndrome. Across wildly different histories, languages and climates, humanity converged on the same folk wisdom: don't be strange, or you'll pay for it.

This wasn't cruelty. It was actuarial advice. For nearly all of human history, being unusual was a losing bet.

The long reign of the average

Consider the payoff matrix your great-great-grandmother faced. Unusually weak? Fewer resources for you when the harvest fails. Unusually strong? Off to a war someone else chose. Unusually clever, or unusually devout, or unusually good at noticing that the local authority was talking nonsense? The stake, the pyre, the exile.

Societies did build escape valves, of course, because a population with zero variance dies the moment the environment shifts. But look at the design of these valves — they're all containment structures. Temporal: carnival, Saturnalia, the Dionysia, three days a year when the rules invert and then snap back. Spatial: the wilderness, the monastery, the ship going west. Role-based: the jester, the holy fool, the hermit. Weirdness was permitted, but only in a designated enclosure, like a fire lit in a stone hearth.

Then came the machines, and with them the metaphor that has organised our institutions for three hundred years: the world as clockwork. And in a clock, what do you want from a cog? You want it round. You want it identical to the cog next to it. You want it replaceable from stock. This is not a conspiracy — it is simply what happens when you try to build reliable complexity out of unreliable parts. You standardise the parts.

We standardised the parts. That's what mass schooling was for, in the plainest possible terms: not to make children stupid, but to make them interoperable. And it worked spectacularly. The cathedral, the railway, the vaccine campaign, the moon landing — none of these are achievable by a society of geniuses who cannot stand in a queue.

So for a very long time, the rational individual strategy and the rational collective strategy pointed the same direction: sand yourself down. Fit the socket. Only those who genuinely could not suppress their strangeness went ahead and paid for it.

What the machines just made cheap

Here is the thing that has quietly changed, and I don't think we've absorbed it yet.

A large language model is, quite literally, a distribution. It is a compressed map of what people typically say, typically write, typically decide, typically produce. Its whole magic trick is prediction. And the corollary of that magic trick is brutal and simple:

If you are predictable, you are already in the training data.

Not metaphorically. Actually. The competent, reliable, on-brief, in-distribution version of your professional output has been absorbed, averaged, and can now be regenerated at the price of electricity. You cannot out-cog a robot any more than a nineteenth-century weaver could out-shuttle a loom. The economic value of being a good cog is collapsing toward the marginal cost of a GPU cycle, and that cost is falling faster than your salary can.

Which means the ancient calculus has flipped. For thousands of years, the safest thing was to be average. Now the average is the most exposed position on the field. The tail is where the shelter is.

Think of it as compression. What is compressible is cheap, because compression is exactly what these systems do. Your value to the world is increasingly a function of how much of you fails to compress — how much residue is left over after the model has done its best to predict you.

Weirdness is not randomness

Now, the obvious objection, and it's a good one: machines can generate noise all day. Crank the temperature up and you get infinite strangeness, instantly, for free. If randomness were the scarce resource, we'd have already solved it.

So let me be precise about what I mean by weird.

Randomness is a random walk — it goes nowhere, it has no memory, and each step contradicts the last. Weirdness is a strange attractor: a trajectory that is unpredictable in its details but unmistakable in its shape. You cannot guess Thelonious Monk's next note. You can identify him in four bars. That combination — locally unpredictable, globally coherent — is the thing that is genuinely hard to fake, because it isn't a property of any single output. It's a property of a life.

And a life is expensive. Real weirdness is paid for — in years spent on an obsession nobody rewarded, in injuries, in embarrassing public failures, in the peculiar wiring you were born with and had to negotiate with daily, in the particular grief and the particular joy that happened to you and to nobody else in that exact configuration. That's the residue. That's the incompressible part. Not the aesthetic of eccentricity, which is very much in-distribution and available in beige — but the actual costly idiosyncrasy underneath.

There's a monoculture warning here too. We already know what happens when a system optimises for one predictable variety: the Gros Michel banana was delicious, uniform, and globally dominant until a fungus took the whole species out in a decade. We are now replanting it with the Cavendish, and the same fungus is coming for that one too. Monoculture is efficient right up until the moment it is extinct. A culture that lets AI iron out its variance is running the same experiment on its own imagination — which is the actual argument behind Artificial Creativity Manifesto.

The conscription

If this is right, quite a lot has to be redesigned.

Education was built to hammer the stakes flat. It now needs to do the opposite without producing monsters — because weirdness without scaffolding, without tolerance, without enough shared ground to make empathy possible, is not a gift, it's a hazard. You need more structure to grow spiky people safely, not less. Vygotsky's scaffolding, deliberately built and deliberately removed.

Community gets harder. The less alike we are, the more work connection takes. The early internet let weirdos find their tribe, which was beautiful; the algorithmic internet then sorted those tribes into legible market segments, which was not. We'll need places designed for cross-pollination between the strange, not just clustering among the similar.

Economy must learn to fund failure. Weird things fail more. That's the deal. UBI might be part of it; tenure suggests that merely removing pressure isn't enough on its own. We'll have to actively reward the long shot.

But let me end on the uncomfortable note, because I don't think this is good news exactly.

"The future belongs to the weird" sounds like a coronation for the misfits. It isn't. It's a conscription notice, and it's addressed to almost everyone. Most of us spent thirty years learning to be a reliable component, and we were right to — it was the correct answer to the question being asked. The question has changed. Now the world wants back the thing we were carefully trained to hide.

The stake that sticks out no longer gets hammered down.

It gets hired.

On the off chance you'd like to subscribe to my writing, please feel free to subscribe on twitter or use this RSS feed. Thank you.