It began, in the public imagination at least, with a silencer and a sidewalk in Midtown Manhattan.
The December 2024 killing of Brian Thompson, CEO of UnitedHealth Group, shot outside a hotel before an investor conference shocked the country less for the act itself than for the reaction to it. Social media didn’t mourn. It deliberated. Memes circulated. The alleged shooter became, in certain corners of the internet, a folk hero. The message scrawled on the shell casings “deny, defend, depose” became shorthand for a rage that had been building for years against a system perceived as automated, indifferent, and untouchable.
Thompson wasn’t an AI executive. He ran a health insurance company. But the public’s response revealed something important: the target wasn’t the man. It was the architecture. The algorithm that denied claims. The faceless process optimized for shareholder return at the cost of human outcomes. The system that couldn’t be argued with, appealed to, or held accountable.
That was the opening act. The attack on Sam Altman’s San Francisco residence in early 2026, an incendiary device, around 4:00 AM, the kind of security breach that makes headlines before quietly disappearing into the news cycle is the second. The pattern is becoming harder to ignore.
We have entered a new phase. The honeymoon is over.
From Insurance to Intelligence: The Same Grievance, Escalating Stakes
The throughline between Thompson and Altman isn’t ideology. It’s alienation. In both cases, the public identified a class of people whose decisions — mediated by systems, scaled by technology, insulated by capital — shape the material conditions of millions while remaining structurally immune to consequence.
The difference is magnitude. UnitedHealth’s algorithm denied claims. The AI architectures being built now are projected to displace, by Goldman Sachs’ estimate, roughly 300 million jobs globally, a figure that looked aggressive when published and increasingly conservative today. When the algorithm takes the mortgage payment, the anger doesn’t target the code. It never does. It targets the architects.
The Architecture of the New Divide
For decades, ‘fortress living’ was considered a symptom of failed institutions — something that existed where governments were corrupt, contracts were unenforceable, and the rule of law stopped at the gate of whoever could afford private security. Students of emerging-market economies recognized the pattern: when wealth concentrates faster than institutions can adapt, the wealthy stop relying on the state and build parallel systems. The state, in turn, stops serving the displaced.
That model is no longer foreign to the West, it is becoming a rational response conditions the West is creating.
The leaders most visibly associated with AI’s advance, Sam Altman at OpenAI, Alex Karp at Palantir, Dario Amodei at Anthropic now operate under security profiles more typical of heads of state than technology executives. This is not incidental. It reflects a public perception that has shifted from viewing these figures as innovators to viewing them as decision-makers whose choices determine the economic fate of millions.
That perception, however distorted in its most extreme expressions, contains a structural truth: the concentration of AGI development in a handful of private organizations represents a novel form of capital accumulation. Unlike Vanderbilt’s railroads or Rockefeller’s refineries, which demanded vast human labor to function, the new infrastructure compounds value while actively reducing its dependence on human input. The owners of the model, in a very real sense, need fewer of us with each iteration.
Hyper-Capitalism’s Logical Conclusion
If capitalism is the disciplined pursuit of efficiency, then artificial general intelligence is its logical endpoint. Efficiency, taken to its absolute form, has no tolerance for the friction of human wages, human error, or human dissent.
What emerges is not a wealth gap in the traditional sense. It is something closer to an existential divergence: a Protected Zone, private enclaves, AI-optimized systems, Palantir-grade security, coexisting with an Automated Fringe, where the gig economy’s “boss” is an unfeeling API and the social safety net is being stress-tested by a deficit that no political coalition seems willing to address.
The 19th-century colonial outpost is the uncomfortable analogy here: a small, highly protected administrative class managing a vast automated estate, with the gates secured against the very population the technology was publicly promised to uplift.
The Question for This Room
The people reading this; technology leaders, enterprise architects, capital allocators, you are not bystanders. The decisions made in the next five years about deployment velocity, workforce transition investment, and the governance of AI systems will determine whether the fortress model is inevitable or a choice.
Brian Thompson didn’t build the algorithm that denied those claims. He inherited a system, optimized it, and answered to a board. Sam Altman didn’t set out to displace the global workforce. He built something that might. The distinction matters legally. It is becoming increasingly irrelevant to the people on the other side of the gate.
The Molotov cocktail is a symptom. The disease is the absence of a credible answer to a straightforward question: In a world where intelligence is a utility owned by a few, what is the social contract with the many?
That question will not wait for the next product cycle. And the people asking it most urgently are no longer doing so with words.
What’s your read? Is decentralized AI governance a realistic counterweight — or are we already past the inflection point?
