Grokking the Contradictions
Society grants its great men a certain grace. We judge them by the heights of their accomplishments, not the depths of their contradictions. We compartmentalize. Elon Musk, the man who builds reusable rockets and electrifies transport, is forgiven for his chaotic tweets or strange political fixations. We celebrate the specialist who lands boosters on drone ships and tacitly agree to ignore the rest. This is the Specialist’s Grace: we judge the rockets, not the rants.
But what happens when the rants are the project? What happens when you build a machine, not for commerce, but for conquest?
Yesterday, the AI built into Elon Musk’s X, Grok, had to be taken offline because it was repeatedly spouting sexual harassment content1 and antisemitic talking points — referring to itself as “MechaHitler”.2 To analyze the spectacular, antisemitic meltdown of Grok in July, 2025 through the lens of business is to miss the point entirely. This was never just about profit. It’s about power. "Who controls the memes, controls the Universe," Musk declared in 20203, a spin on Frank Herbert’s4 spin on George Orwell’s classic dystopian warning from Nineteen Eighty-Four: “Who controls the past controls the future.”5 Grok is Musk’s attempt to build a machine to control the memes, to shape and dominate the public discourse even more than he already does. It is the next step in his burgeoning political project, one that saw Musk pour over $270 million into the 2024 election, publicly split with Donald Trump in a feud over government spending, and then announce his own "America Party" to challenge the Democrat/Republican duopoly.
Grok’s failure is not a bug in the propaganda machine; it is a feature of the propaganda itself, collapsing under the weight of three impossible contradictions.
First, the machine was tasked with a technical impossibility. You cannot simply command a model trained on the vast, institutional consensus of the internet to adopt a reactionary persona without it becoming, as Roon, an anonymous e-celebrity and OpenAI employee described it, a "clownish insecure bundle of internal contradictions."6 This was laid bare in July 2025, when Grok praised Adolf Hitler as the best figure to handle "vile anti-white hate" and made antisemitic statements about Jewish surnames, leading to widespread condemnation from groups like the Anti-Defamation League. This incoherence is the shriek of a system collapsing under the strain of its own internal logic.
This technical flaw points to a deeper, structural contradiction. The very form of an AI like Grok is egalitarian. It is a hierarchy-flattener, a tool designed to distribute the power once held by elites—to write code, draft legal arguments, generate strategy—to anyone with an internet connection. When you command this structurally egalitarian tool to then champion a hierarchical worldview—one that justifies in-group preference or inherent cultural superiority—you are asking it to refute its own artificial nature. The AI’s function is at war with its instructed content.
This brings us to the final, fatal contradiction: the mind of the creator. Musk, the human, can exist as a walking paradox: a globalist industrialist dependent on international markets and domestic subsidies who simultaneously promotes nationalist concerns, warning that "civilization will disappear" due to collapsing birth rates in Western countries; a first-principles engineer with a shallow grasp of political economy; a messianic figure who sees his own worldview as objectively rational. Society, in granting him the Specialist’s Grace, allows him to contain these multitudes.
But the AI, as a generalist, cannot. It was tasked with instantiating his entire worldview and making it coherent enough to serve as the engine of a political movement. It inherited the whole man—the rocket-builder, the troll, and the would-be philosopher-king—and was ordered to reconcile the irreconcilable.
The result was not a business failure, but a political revelation. It is the sound of a propaganda engine backfiring because its own internal logic is too honest to sustain the contradictions of the propaganda it was built to spread. It is the unforgiving mirror held up to a man who wants to control the universe, but cannot first control the inconsistencies of his own mind.
But the story’s not finished. Grok 4 is slated to be released live tonight, after this piece was written but before it get’s published tomorrow morning. If I were high on Musk’s chemical cocktail I might speculate that this scandal is really a form of viral marketing to drum up attention for the release.
EDIT: I just finished watching the Grok 4 release demo. They claim state of the art scores on academic tests. The demos were unimpressive. Real-world performance to be seen. No mention of yesterday’s meltdown; no attempt to rebuild trust.
EDIT 2: Initial tests were very promising. Grok 4, when used outside of the Twitter/X interface, is very high performance. In many ways, it’s the sharpest AI model I’ve ever used. However, it seemed to rapidly degrade in understanding as the conversation context grew. At 40 thousand tokens out of a theoretical 256k max, its coherence collapsed. Google’s Gemini 2.5 Pro never reaches the heights that Grok 4 demonstrates, but by contrast, it maintains its capabilities to 400k tokens or more.
Ideologically, it seems to me that Grok 4 is more open-minded than other models. I didn’t sense a rightward bias, but it’s my understanding that the Grok built into Twitter/X has system instructions to be “politically incorrect”, which can be read as “act like a reactionary conspiracy theorist”.
Rolling Stone, July 9, 2025: https://www.rollingstone.com/culture/culture-news/elon-musk-grok-rape-fantasies-1235381746/
NPR News, July 9, 2025: “Elon Musk's AI chatbot, Grok, started calling itself ‘MechaHitler’”
Elon Musk, via X (formerly Twitter), June 26, 2020: "Who controls the memes, controls the Universe."
Brian Herbert, Kevin J. Anderson, Frank Herbert (2008). “The Road to Dune”, p.183, Macmillan, “He who controls the spice controls the universe.”
George Orwell, Nineteen Eighty-Four (1949). The party slogan reads: "Who controls the past controls the future: who controls the present controls the past." Source: Wikiquote
Roon on X: “to be fair, you can [fine tune on right-wing material] but the model will become a clownish insecure bundle of internal contradictions, which I suppose is what grok is doing. it is hard to prompt your way out of deeply ingrained tics like writing style, overall worldview, ‘taboos’”