You look at the headlines, shake your head, and think, “How is Washington so easily duped by Silicon Valley?” You see politicians bragging about the latest AI deployment, and you assume they just don’t understand how large language models work. You’re wrong. They aren’t naive. They know exactly what they’re buying. Washington officials aren’t ignorant; they’re just buying the performance of competence with our money.
We’ve all seen the press releases. A frontier AI lab “deploys” a system for a government agency. It sounds impressive. America is leading! But if you look at what’s actually being procured, it’s mostly vaporware, basic query automation, and fragile workflows that collapse the second they face real-world complexity.
We love to blame “gullible” politicians. We mock the bureaucrats who don’t know their prompts from their parameters. But that lets the real culprits off the hook. It assumes the system wanted to succeed and just failed. The reality? The failure is perfectly acceptable. In fact, it’s the feature, not the bug.
Think about the political incentives. If a senator approves a $50 million AI contract and it turns out to be useless trash, what happens? Does the senator get primaried? Do they lose their chairmanship? No. They go on TV and say, “I worked to make America ready for the AI revolution, and the private sector failed to deliver.” Either way, the senator gets the political capital of being an “AI leader.” And that $50 million? That’s your money. When the hype secures votes and budgets, nobody demands proof of capability.
This isn’t a market failure where a buyer gets tricked by a seller. This is a perfectly calibrated system where the seller gets astronomical valuations, the buyer gets political cover, and the public gets drained. As one sharp observer noted about this exact dynamic: “They are getting paid with our money.” That is the entire business model.
The frontier labs don’t need to deliver a working product that actually solves bureaucratic bloat or improves governance. They just need to deliver enough of a narrative to keep the press releases flowing. The actual experts in these agencies—the civil servants who use LLMs daily and know exactly where they break down—are ignored because their warnings contradict the political narrative.
The procurement structure isn’t designed to find the best tools; it’s designed to subsidize the loudest stories.
Stop mocking Washington for “not getting” AI. They get it perfectly. They are pouring public funds into a speculative bubble because the bubble benefits everyone at the negotiating table. The problem isn’t Silicon Valley hucksters meeting Washington fools. The problem is an incentive structure that makes performative innovation more profitable than actual problem-solving. The real fools aren’t the ones buying the garbage in Washington. The real fools are the ones paying for it.
FAQ
Q: Aren't some of these AI systems actually useful for government tasks?
A: Some are, but the scale of these frontier contracts is based on hype, not proven ROI. The issue isn't whether AI can summarize a document; it's that billions in public money are flowing based on narrative strength rather than verified capability.
Q: What does this mean for the average citizen?
A: It means your taxes are subsidizing Silicon Valley valuations while critical public infrastructure decays. You are paying for PR stunts that fail silently, while the real governance problems these systems were supposed to solve remain completely unaddressed.
Q: So you're saying politicians and AI labs are actively conspiring against the public?
A: Not a conspiracy, just perfectly aligned incentives. Labs want high valuations; politicians want 'innovation' credentials. Neither is punished when the AI underperforms. It's a rational collusion where the only party bearing the cost is the taxpayer.