I was walking through the Mission District last week after a frankly tremendous matcha meeting with a seed-stage founder who is disrupting the concept of the weekend, when I decided to drop by OpenAI’s headquarters. They were in the midst of setting up what they called an "election war room," a dedicated, multi-screen command center designed to fight the absolute tsunami of synthetic disinformation currently threatening the midterm elections. Most legacy pundits are wringing their hands about the death of objective reality and the collapse of the American democratic experiment. But as I watched a proprietary dashboard tracking millions of perfectly hallucinated deepfakes generated by their own large language models, I felt a profound sense of awe at the sheer scale of the total addressable market they have unlocked.
The best founders I know understand that true market dominance doesn't come from finding a niche, but from creating an entirely new ecosystem of demand. For years, Silicon Valley has struggled with the messy, low-margin business of content moderation. What OpenAI and their rivals at Anthropic have realized is that if you want to monetize the cleanup, you first have to industrialize the spill. By simultaneously deploying the greatest automated-bullshit generators in human history and then establishing bespoke war rooms to heroically monitor the resulting chaos, these companies have achieved a level of vertical integration that would make John D. Rockefeller weep.
Inside the facility, the synergies are palpable. On one floor, server racks are burning through gigawatts of power to help anonymous users generate photorealistic images of California Governor Gavin Newsom committing arson. On the floor directly above, a team of trust-and-safety engineers is bravely utilizing those exact same computing resources to identify and flag the arson photos as synthetic. It is a breathtaking closed-loop economy. They are the manufacturer, the distributor, the crisis, and the highly-paid crisis consultant all rolled into one heavily optimized corporate structure.
Consider the sheer operational efficiency of an AI-generated scandal in a battleground district. In the past, manufacturing an October surprise required opposition researchers, private investigators, and a sympathetic media apparatus. It was artisanal. It didn't scale. Today, a single prompt engineer in a basement can A/B test ten thousand distinct variations of a video supposedly showing Texas Governor Greg Abbott endorsing a ban on oxygen. The war room monitors which of these synthetic narratives achieves the highest click-through rate, providing invaluable data back to the core engineering team about human cognitive vulnerabilities. It is the ultimate feedback loop. They are utilizing the midterms as a massive, unpaid beta test for the underlying architecture of mass persuasion.
The basic business mechanics of the midterm disinformation cycle are flawlessly designed. A bad actor pays a standard API subscription fee to access GPT-4, using it to spin up ten thousand localized blog posts claiming the polling stations in Maricopa County have been relocated to the bottom of the Grand Canyon. This is pure, high-margin software revenue. Then, philanthropic organizations, campaigns, and tech watchdogs raise millions of dollars to combat this exact narrative, utilizing enterprise-tier API calls to those same AI models to parse and filter the noise. OpenAI gets paid to print the counterfeit currency, and they get paid again to sell the counterfeit detectors.
We view the structural integrity of the American republic not as a vulnerability, but as a dynamic, high-engagement platform where both the generation of epistemic ruin and the subsequent mitigation of that ruin can be aggressively monetized.
The fossil fuel industry never figured this out. Standard Oil had to drill the crude, refine it, and ship it, but they never quite managed to charge municipalities a monthly fee for the carbon scrubbers. OpenAI has leaped past the extraction industry entirely. They are selling the digital toxic waste, and then turning around and selling the hazmat suits to the exact same electorate, booking revenue on both sides of the ledger.
Anthropic, to their credit, is taking a slightly different but equally brilliant approach with their Constitutional AI. They have positioned themselves as the premium, ethical alternative, the Whole Foods of automated psychological warfare. Their war room is less about frantic mitigation and more about demonstrating that their proprietary models can selectively pause the erosion of objective reality if the right enterprise contracts are signed. It is the freemium model applied to the very concept of empirical truth. You can have the baseline democratic chaos for twenty dollars a month, but if you want the version of the internet where your grandmother isn't radicalized by a synthetic video of New York Mayor Zohran Mamdani eating a golden retriever, that requires a bespoke enterprise SLA.
The real innovation isn't the large language model itself, it's the realization that democracy is a legacy system with incredibly weak firewalls, making it an ideal environment for our enterprise resilience products.
I discussed this over a tasting menu in Atherton with a prominent venture capitalist, and we both agreed that the traditional media sector has fundamentally misunderstood the product lifecycle here. Newspapers and cable networks spent decades trying to organically aggregate human outrage, a labor-intensive process reliant on actual events happening in the physical world. Generative AI has abstracted the outrage layer entirely. The war rooms aren't there to stop the disinformation; they are there to regulate the flow, ensuring the American electorate is exactly agitated enough to keep scrolling, but not so agitated that the electrical grid goes down and takes the data centers with it. It is demand-side management at its finest.
What we are witnessing in San Francisco right now is the birth of the B2G—Business to Government—extortion model, and frankly, it is overdue. Why should tech companies provide baseline geopolitical stability for free? By demonstrating that they hold the absolute power to flood a Tuesday midterm with an impenetrable fog of synthetic madness, OpenAI and Anthropic are establishing leverage. The war rooms are just showrooms. They are inviting lawmakers in, pointing to the screens flashing with thousands of fake audio leaks and deepfaked candidate breakdowns, and softly suggesting that keeping this contained might require favorable regulatory frameworks moving forward.
In this economy, you have to admire a leadership team that refuses to leave money on the table. The hand-wringing over democracy is a distraction from the fundamental bull case: truth is no longer a public utility, it is a subscription service. And as long as these companies own the only viable factories for both the disease and the cure, I will be buying the dip.