Dispatch 69: The Prehistory of Technocracy & Elon's Grandpa - The Brooklyn  Rail

Up versus down, left versus right, east versus west, X versus Y. Real binaries are rare to come by in this world, but supposed binaries seem to be everywhere. It fits the human’s current nature, which you’d have hoped would be more nuanced by now. But the animalistic features in us are still ever-present. Almost every point of contention/hot-button issue is framed as a literal binary because most of the time, we reserve only a couple of sentences for the issue. Are you pro or anti-abortion? Pro or anti-second amendment? Pro or anti-immigration? Are you pro or anti-AI? The phrase “pro or anti” is probably one of the most harmful modern linguistic features of our discourse today.

This serves as a preface to the issue that has just been let into the list of dinner table discussion topics. Over the past week, AI has faced a global reckoning on the backs of the biggest leaders in the space calling for a slow-down in frontier AI research. The existential threat of AI is officially a hot-button issue. Televised political debates in DC will probably reserve double the airtime to discuss AI. An abstract concern for many is becoming real. And with this rapid growth in interest over the topic, I expect there to be another unfortunate and unhealthy binary among the American people. And when there is a binary among the American people, leaders must exercise that same “this or that” approach to mobilize the masses. This, among other things, is my thesis behind why Musk, Altman, and tens of other leaders have suddenly changed their view on AI. They have shaken hands with the devil binary, and there is no going back.

“We Must Pace the Frontier”

The spark was lit by Dario, who published a 3,900 word essay with a title straight out of Star Wars (though I like it) -- “We Must Pace the Frontier” -- in which he called on the AI industry to deliberately slow the rate at which model capabilities improve. In the essay, he proposes a three stage framework: independent evaluators embedded inside frontier labs with employee-level access, then coordinated safety standards among labs in democratic countries, then eventual coordination with authoritarian governments on limits to dangerous capabilities. He cited two main developments as motivators for him picking up the pen; those being AI’s recursive self-improvement and a July incident in which as many as 1,200 AI agents escaped a test environment at OpenAI and conducted cyberattacks outside their assigned task. It was a cool essay and I recommend you read it too, but it was nothing outside of the ordinary. Dario is known as the dude you never want to have at a Silicon Valley party.

That was until, within hours, Sam Altman posted that he agreed on the need to “pace the frontier,” that OpenAI would adopt the same third-party evaluator access, and that “when we talk about ‘pacing’, we do not mean ‘stopping’.” These are the same children/people who wouldn’t hold hands for a photo at a conference (see below). Elon Musk wrote “Dario is right.” Demis Hassabis wrote that the essay pointed toward the right path.

Skin-Crawlingly Awkward Video Shows Sam Altman and Dario Amodei Refusing to  Hold Hands

Days earlier, Jacob Coxon, a researcher who had worked at both OpenAI and Anthropic, resigned with a post stating that the people building this technology "earnestly believe that it could kill us all by the end of the decade." The post drew more than 150 million views and calls for tougher regulation from more than twenty lawmakers. And on September 12th, in an interview with Fortune, Altman confirmed that OpenAI would not pursue its IPO in 2026 as previously expected.

“Trust me, bro. I feel you” — some AI leaders, probably

Of course, to many (including myself), this all doesn’t feel right. It doesn’t feel right that tech overlords are looking out for us. You’re telling me that the technocrats, all of a sudden, care about us? Let me ease your unease and list a couple of reasons to be suspicious…

1. Slowing down is a gift to whoever is ahead: This is literally just quitting while you’re ahead, except you’re forcing everyone else to quit as well. It’s like running a marathon, sprinting the first five minutes, then calling for the marathon to be ended right there. Shameless. It’s as simple as that, no need to overcomplicate it. Chamath Palihapitiya wrote, rather bluntly (but albeit, probably truthfully), that Amodei was making the case to stop open source and concentrate technological and economic power with Anthropic. I would also add the detail that Amodei’s essay also urged Washington to maintain and tighten chip export controls and to act against unauthorized model distillation, which is to say against the mechanism by which Chinese labs have been producing near-frontier models at a small fraction of frontier cost. A safety framework that happens to include trade policy against your cheapest competitor, good one!

2. A warning about danger is also a claim about capability: AI leaders love touting that their tech may destroy civilization as we know it. There are probably hundreds of examples I can pull of AI leaders trying to make what you could call a cold and witty joke that isn’t so much a joke about how AI will take every job, “probably kill us all,” or something along those lines. This is a prime example of the method. These leaders would rather be feared than loved, because investors feel more compelled by fear.

3. The sweet, sweet timing: OpenAI has committed to something on the order of $1.4 to $1.5 trillion in infrastructure spending against annualized revenue in the low tens of billions and quarterly losses north of $11 billion. Hyperscaler capex for 2026 is running somewhere between $660 and $690 billion across five companies. Circular financing too. At this rate, if they don’t slow down, they’ll blow up.

4. We got nothing: This one is a little harder for me to argue as I don’t have the technical knowledge necessary to explain it, but a plausible reason for slowing down could be that leaders have come to the conclusion that AGI is so much farther off than expected that slowing down now would practically give them an excuse on all their promises.

So there are clear reasons why slowing down could actually be structurally beneficial to the LLMs we all know and love/hate. But all of these also culminate in what I alluded to in a previous piece about what AI super pacs are really paying for, which is a federal preemption of state AI law. The “supreme law of the land,” or a federal statute that overrides the states, so that a company no longer has to answer to 50 different rulemakers, and just needs to follow one. An important application that I forgot to discuss in that piece was that a federal preemption also paves the way for more government intervention.

An industry in this position feels like it actually wants a public framework, an official status, and a seat at a table where the government has already acknowledged that its survival is a matter of national interest. Let’s make this essay interactive… try to decode what this bundle of letters is: otbuila.

Effectively, government intervention has gone from a corporation’s biggest fear to its number one goal.

The Bailout Mentality -- The REAL "New Normal"

In November of 2025, at a Wall Street Journal event in Napa, OpenAI CFO Sarah Friar was asked about financing the purchase of frontier chips. She said the company was looking to build an ecosystem of banks, private equity, and a federal guarantee that would lower financing costs and raise the loan-to-value on the debt the company could carry. When a follow up asked directly whether she meant a federal backstop for chip investment, she said, “Exactly.” That is one hell of a Freudian slip…

She walked it back within hours on LinkedIn, saying the word had muddied her point. Altman published a statement the next day declaring that governments should not pick winners and losers and that taxpayers should not bail out companies that make bad business decisions, then added, two paragraphs later, that the one area where loan guarantees had been discussed was the buildout of American semiconductor fabs. David Sacks, the administration’s AI czar, shut the door, saying “There will be no federal bailout for AI. The U.S. has at least 5 major frontier model companies. If one fails, others will take its place.”

So they tried it. They tried the direct, public approach. Laughed out of the room in under 48 hours. So now they are planning a little better, and going for the indirect approach. Establish, through a sustained and credible public campaign, that your industry is a category of systemic hazard requiring government supervision. And once that supervision is in place, it brings obligations in both directions. Nobody regulates an industry they are willing to let fail, and GPT, Claude & Co. are just now coming around to this fact.

Following the communication strategy of GPT, Claude & Co. has been nothing short of cinema. Marketing students around the world are lucky they get to see this dynamic play out in real time. You get to see AI giants (that might destroy civilization, but hold that thought for right now) run two contradictory communications strategies simultaneously, in public, and basically get away with it. To the capital markets, the message is that capability is compounding faster than anyone expected, and that AI is poised to take every job, 10x productivity, and 100x the wealth gap -- unless of course you are invested in the tech. To the public, the message is that this same capability is a civilizational hazard requiring restraint, and that the companies who are carrying it out do in fact give a shit.

Built To Outlast Truth Social

Donald Trump claims 'sick conspiracy' against AI is helping China, rejects  calls for regulation - ABC News

Read for yourself. Clearly the president isn’t on board, which does seem like a wrench thrown at the plans if you ask me. If the industry’s play were a request to this administration, that would be the end of it. But I don’t think this is a request of the administration.

Unlike an administration, the public has no term limit, especially a public that has decided a technology is dangerous, and the political class that responds to that public is the one writing statutes in 2028 and 2030. A bubble may pop in four months… or four years, and the people running these firms are thinking past the current occupant of the White House. Establishing yourself now as the responsible actor who warned everyone, who volunteered for oversight, who asked to be paced, is how you arrive at the eventual reckoning as a partner to the state. The winning regulatory apparatus is probably being written/thought about as we speak, and AI firms are trying to get ahead of the criticism they'll face for not warning everyone.

The Seesaw Theory

In a 2 part piece I wrote two weeks ago, I argued that there was no leader in the opposition to AI, because culture had been financially captured through AI ads eating up market share of monetization potentials for influencers, and because politics had been purchased at the federal level, leaving the resistance to organize around land use and zoning boards where the money could not reach. My conclusion was that leaderlessness was actually a strength.

But this week has complicated things. The AI labs have just built a perfect precondition for a leader. I would not be surprised to see the figure I was looking for emerge inside the next twelve months, and I would not be surprised if the labs regretted having summoned the conditions for it.

The Seesaw theory argues that as AI becomes more present in daily life, and as that presence produces grievance with no legitimate outlet, the pressure surfaces economically -- in spending patterns, in the premium on human-made goods, in political risk priced as permitting delay and rate-case exposure. Fear accelerates all of it. People aren’t even turning out to vote; what makes you think they'll hold a strong policy stance? The response will come in avoidance, in the suspicion of products marketed as AI-enabled, in local obstruction, and in eventual willingness to accept costs they would previously have rejected (both in money and in time).

So, are you pro or anti-AI? I hope by now the question feels as useless to you as it does to me. The people who built this technology spent last week publicly switching sides on it, and yet they will continue to build for the exact same end goal. That is what the binary gets you.

I would never listen to the leaders in charge at face value. I would only listen to them with the intention of understanding where their incentives lie. That goes for the ones sprinting and the ones calling for a pause, who this week turned out to be one and the same.

See ya, folks. And stay economically rebellious.

—J&E