AI Is Not Going to Save You (But It Might Help)

There’s a scene playing out in boardrooms across the country right now. Some VP of Innovation—a title that didn’t exist eight years ago—is standing in front of a slide deck with the word ‘AI’ on it seventeen times. The slide deck was probably written by ChatGPT. The VP is sweating slightly. The board is nodding. Nobody is asking the obvious question.

The obvious question is: What problem are we actually solving here?

We are three-plus years deep into the most aggressively marketed technology cycle since the dot-com bubble, and the gap between what AI is being sold as and what it actually does has somehow gotten wider, not narrower. Let’s talk about that gap. Honestly. Without a pitch deck.

The Hype Machine and How It Works

Here’s the thing about hype cycles—they’re not accidents. They’re business models.

When Microsoft, Google, and OpenAI tell you that AI is going to revolutionize everything from healthcare to your morning email, they’re not wrong exactly. But they’re also not just sharing a vision. They’re selling cloud compute. Every token you generate runs on their servers. Every API call is a billing event. The revolution they’re describing happens to be enormously profitable—for them.

And just this past week, Microsoft had to shut down over 70 of its own GitHub repositories after hackers pushed malware targeting users of AI coding agents—tools designed to steal credentials from developers trusting AI to write and manage their code. Think about that architecture for a second. We’ve built systems where an AI agent has enough access to your environment that compromising it can drain your credentials. That’s not a bug. That’s the design of maximum dependency.

“The question isn’t whether AI is useful. It clearly is. The question is: useful to whom, and on whose terms?”

The hype machine works by conflating real capability with imagined capability, then letting your own imagination do the rest of the marketing. You fill in the blanks. You picture the version that solves your specific problem perfectly. Then you buy the subscription.

What AI Actually Does Well (And What It Doesn’t)

Look, I’m not here to tell you these tools are worthless. That would be dishonest and also wrong. I use them. They save me time. But there’s a meaningful difference between a useful tool and a sentient oracle.

AI language models are extraordinarily good at a specific category of tasks: pattern completion at scale. Writing a first draft? Great. Summarizing a document? Excellent. Helping you debug code by recognizing patterns from millions of similar problems? Genuinely impressive. These are real things with real value.

What they’re not good at: novel reasoning under uncertainty, knowing what they don’t know, staying accurate on fast-moving topics, or anything that requires actual accountability. They hallucinate with complete confidence. They’ll cite papers that don’t exist. They’ll give you the statistically likely answer, not the correct answer—and they can’t always tell the difference.

The Hollywood conversation is a good case study. AI in filmmaking—voice synthesis, visual effects, script analysis—is genuinely changing production economics. But the idea that AI is going to replace the human storytelling impulse? That’s a pitch for nervous executives, not a technical reality. The tools are real. The narrative around them is marketing.

The Surveillance Tax Nobody Talks About

Here’s what gets buried in the AI conversation: the data infrastructure being built to power it isn’t neutral. It has opinions about who owns your information.

A company called SignalTrace just rolled out a product that links Bluetooth devices—your phone, your AirPods, your smartwatch—to license plate readers. The system correlates devices that travel together, building behavioral profiles. This is being sold as a law enforcement tool. It rides on the same data economy that funds AI development: mass collection, centralized storage, algorithmic processing.

And separately, a farmer in Texas donated 87 acres to his city in 1999 to be used as a park. The city just sold it to a data center developer for $10 million. That’s the physical infrastructure of the AI economy: land that was supposed to be a commons, converted into server farms, processing your queries, training on your data, generating revenue for people who weren’t even in the room when the original promise was made.

“Every data center is a bet that your attention and your data are more valuable than whatever else that land could have been.”

The surveillance angle isn’t paranoia. It’s just following the money. AI requires data. Data requires collection. Collection requires infrastructure. Infrastructure requires land, power, and your continued participation. Who benefits from that chain? That’s the question the hype cycle really doesn’t want you sitting with.

How to Think About AI Without Losing Your Mind

The useful frame here isn’t “AI is amazing” or “AI is fake.” Both of those are lazy. The useful frame is: What is this tool actually good at, and who controls it?

Use AI for leverage, not dependency. There’s a massive difference between using a tool to amplify your own thinking versus outsourcing your thinking to a tool. The first makes you better. The second makes you replaceable—by the tool itself, or by whoever controls it.

Get curious about self-hosted options. Ollama, LM Studio, and open-weight models like Llama and Mistral have gotten genuinely good. Running a model locally means your queries don’t train someone else’s product. It means your data doesn’t leave your machine. It’s slower, it’s less convenient, and it’s worth considering for anything sensitive.

Stay skeptical of AGI timelines. The people giving you specific dates for artificial general intelligence are either fundraising or genuinely overconfident. Prediction markets and serious researchers disagree wildly on this. When someone tells you AGI is two years away, ask yourself who profits from you believing that.

Understand the incentive structure. OpenAI, Anthropic, Google, Microsoft—these are not charities. They are in the business of making AI seem indispensable. That doesn’t mean their products are bad. It means their marketing has a direction, and that direction is toward maximum dependency on their platforms.

The Real Question

Here’s what I keep coming back to: every technology wave in history has been sold as liberation and delivered something more complicated. The internet was supposed to democratize information. It did that. It also built the most effective advertising and surveillance infrastructure in human history. Both things are true.

AI is doing the same thing. It is genuinely useful. It is also being used to build systems of dependency, surveillance, and control that most people aren’t thinking carefully about. The question isn’t whether to use the tools—of course use the tools. The question is whether you’re using the tools or whether the tools are using you.

That distinction is worth a lot more than any AI certification, any vendor pitch, or any slide deck with the word ‘revolution’ in it seventeen times.

So what are you actually building with it—and who owns the infrastructure underneath?