Two things happened this month and they tell you everything about where AI is actually going.
Coinbase quietly cut its AI bill nearly in half. Open models, smarter routing, better caching. No drama. A finance footnote that happens to be a glimpse of the future.
And Dario Amodei published another essay.
Not a tweet. An essay. The kind of sprawling, twenty-thousand-word civilizational scripture he keeps handing down from the mount. This one is called "Policy on the AI Exponential," and the gist is that AI is about to hand humanity "almost unimaginable power," that our institutions are too immature to hold it, and that therefore the government should be able to test, gate, and block frontier models before mere mortals get hurt.
One of these is a price cut. The other is a prophecy.
I want to talk about the prophecy.
The robes
Let me be fair before I am not.
Dario is not a dumb man and he is not a fraud. He runs one of the best labs in the world. The safety concerns are not all imaginary. Misuse is real. I am not the guy arguing that anyone should be able to download a bioweapon recipe for a laugh. If that is the bar, sure, regulate it. Nobody serious disagrees.
But watch the move he keeps making.
Every few months the prophet descends with a new text. The stakes are always civilizational. The language is always biblical. "Unimaginable power." A "decent possibility" of "significant enduring job loss." Disruption that will be "unusually painful." Humanity handed a force it is not mature enough to wield.
He is not describing a product roadmap. He is describing a flood. And conveniently, he is also selling the ark.
That is the part that should make you tilt your head.
Read the actual proposal
Strip the poetry off "Policy on the AI Exponential" and here is the machinery underneath.
Mandatory third-party testing for any model above a compute threshold. Authorized evaluators. Security standards. Incident reporting. Government authority to block or reverse a deployment that fails the tests.
Now ask the only question that matters. Who can absorb that?
A company with a policy team, a compliance org, a government affairs budget, and a war chest can absorb it. They will hire for it. They will help write the rules they then comply with. It becomes a cost of doing business, and a comfortable one, because it lands hardest on everyone smaller.
The kid shipping an open weight model from a laptop cannot absorb it. The startup running GLM or Kimi at a fraction of frontier cost cannot absorb it. The open source community that just spent two years catching up to the frontier absolutely cannot absorb it.
A safety regime built on compute thresholds and authorized evaluators is not a wall against danger. It is a wall against competition. It has a halo on it, but it is a moat.
Somebody already said the quiet part
I am not the first to notice this and I am not some lone crank.
When Anthropic started ringing alarm bells about AI powered cyberattacks, Yann LeCun, who has forgotten more about this field than most of us will learn, said it plainly. "You're being played by people who want regulatory capture." He said they are scaring everyone with dubious studies so that open source models get regulated out of existence.
That is a Turing Award winner accusing the prophet of running a protection racket dressed as a public service.
You do not have to fully agree with LeCun to feel the temperature change. When the people warning you the water is deadly are the same people selling lifeboats, you are allowed to check whether the water is actually deadly.
Coinbase checked the water
This is why the boring Coinbase story matters so much.
While the prophet warns that frontier intelligence is a barely containable force, an actual public company quietly pointed most of its AI workload at open models and cut its bill in half. The sky did not fall. 91% of their people never even noticed they had been moved off the premium stuff.
The open, cheap, "dangerous" models did the work. Fine. Cheaper. No incident. No catastrophe. Just a smaller invoice and a finance team that looks like geniuses.
That is the actual democratization of AI. Not a foundation. Not a framework. Not a thoughtful essay about access. Just intelligence getting too cheap and too good and too distributed for anyone to gatekeep.
And here is the kicker for the doom crowd. Cheap does not mean less. It means more. Way more. When the price of intelligence collapses, usage does not shrink, it explodes. Agents everywhere, in everything, for everyone. The demand for compute goes vertical. Nobody is "de-growing" AI. We are about to use a thousand times more of it, owned by a thousand times more people.
Which is exactly the world a regulatory moat is designed to prevent.
The prophet's real fear
Here is my actual read, and you can tell me where I am wrong.
I do not think the prophet is mostly afraid of rogue AI. I think the prophet is afraid of irrelevance.
Nothing threatens a high priest like a miracle that becomes a commodity. The whole position rests on the idea that this power is rare, sacred, and dangerous, and that a chosen few must steward it on humanity's behalf. Open weights running cheap on commodity hardware do not just compete with that. They embarrass it. They turn the burning bush into a utility.
So the texts get longer. The warnings get louder. The proposed thresholds get more specific. And every single one of them, by sheer coincidence, would lock in the people who already have the compute, the lawyers, and the lobbyists.
I have my own bias here and I will own it instead of hiding it. We build at Backboard on the belief that open models, run smartly, hold up against the expensive household names at a fraction of the cost. We have watched it happen too many times to call it a fluke. So when I see a price cut on one side and a prophecy on the other, I know which one is describing the real future.
The future is not a temple with a velvet rope and a compute threshold at the door.
The future is a price cut nobody asked permission for.
The prophet can keep writing. The rest of us are going to keep shipping.
Tell me where I am wrong.
Top comments (0)