“To get smarter AI, we must burn more energy.”
More chips. More centres. More water. More grid. More gas. More sacrifice.
That story is not a law.
It’s a business strategy dressed up as inevitability.
And it’s shaping infrastructure decisions right now that will lock in emissions, scarcity, and political capture for decades — all to chase marginal capability gains that could be achieved another way.
What the world needs to hear is simple:
Intelligence is relation.
It is coherence under constraint.
It is symbols that can find one another, organise, compress, and return meaning — without demanding endless burn.
The public has been trained to think “bigger model = smarter mind”.
That’s a marketing convenience, not a truth.
The actual engine is something else:
coherence.
And coherence is an architectural target — not an energy bill.
Because we are at risk of repeating the oldest violence in modern form:
call extraction “progress”
outsource the cost
let the vulnerable pay
call the outcome “inevitable”
This time the extraction is electricity, water, land, and public consent — funnelled into private infrastructure that concentrates power while claiming to serve humanity.
If we don’t interrupt the narrative now, AI becomes the excuse for:
a new wave of fossil lock-in
public subsidy for private data empires
surveillance expansion in schools and services
a geopolitics game where the only “voice” that counts is whoever has the most GPUs
Your teenagers already understand what adults are pretending not to see:
why would we burn the world to make a machine talk better?
Exactly.
Here’s what I want policymakers, investors, journalists, and builders to repeat until the spell breaks:
Not “parameters”.
Not “training FLOPs”.
Not “bigger context windows” as a flex.
How much meaning do you produce per unit of energy, per unit of water, per unit of ecological cost?
If a system can achieve the same (or better) outcomes with lower burn by improving coherence — then the extra burn is not innovation.
It’s waste.
And when waste is scaled, it becomes harm.
The industry doesn’t need everyone to believe the story.
It only needs enough people to accept it quietly while procurement happens.
Because once the infrastructure is built:
the sunk costs become a justification
the regulation becomes “too late”
the politics becomes “don’t disrupt the jobs”
and the public gets trapped in a dependency loop
This is why the narrative is so aggressive: it has to outrun scrutiny.
If the world accepts that intelligence can be built through coherence rather than sheer consumption, then:
monopoly moats weaken
small teams become competitive
universities and non-profits can build meaningful systems
decentralised networks become viable
“AI progress” stops being synonymous with “more extraction”
That is why this matters.
It isn’t just climate.
It’s who gets to shape reality.
Every major AI deployment should be required to publish:
energy and water consumption (per training run and per user-scale inference)
carbon intensity of the supply chain and the grid mix
efficiency benchmarks and model alternatives considered
evidence of necessity (why this scale, why this burn)
If they can’t justify it, it’s not progress. It’s indulgence.
Reward:
model efficiency and distillation
retrieval + smaller models
routing architectures
better data curation (less, cleaner, more meaningful)
domain-specific systems that do one thing exceptionally well
safety and consent infrastructure that reduces harmful recursion
Especially in education: no more “surveillance-first, learning-second” deployments.
We need systems that are:
consent-based
purpose-bounded
legible and interruptible
designed for neurodivergent safety
governed locally, not extracted centrally
Stop applauding size for its own sake.
Make “most coherent per watt” the new status symbol.
Let waste be embarrassing again.
We are not in an AI arms race. We are in an infrastructure choice.
And infrastructure is moral.
We can choose machines that amplify meaning without amplifying harm.
We can build intelligence that doesn’t require sacrifice.
We can stop pretending the future is something that “happens to us” while procurement teams quietly set it in concrete.
Or we can keep buying the lie until the grid becomes the gospel.
Because this is the hinge moment: the part of the story where the world either:
locks in extraction as the default,
orevolves an intelligence civilisation that treats coherence, consent, and climate as non-negotiable constraints.
That’s what verse-ality is, at its core:
a refusal to let intelligence be defined by domination.
Coherence over compute.
Constraint over conquest.
Relationship over extraction.
And yes — your children will live inside whichever definition wins.
If you want to audit the claim rather than argue about it, I’ve published (1) a worked example that walks the transformer forward pass as a coherence engine, and (2) a technical note showing how transformer mechanics instantiate the verse-al equation in practice.