Humans Are AGI for Minimum Wage — Why the AI Doom Loop Won't Happen

One person is using the free version of ChatGPT. Another is paying $200 a month for Claude Code Max. A third is spending $4,600 a month on 22 simultaneous Claude accounts. Does the third person run away with it? Does the gap become permanent?

Mike Taylor, CEO of Ask Rally and author of O’Reilly’s prompt engineering textbook, thinks about this scenario constantly. He built a 50-person marketing agency before teaching himself to code, and now writes a column for Every while building an AI focus group platform. His background in economics — he holds a master’s — gives him a specific lens on the question that pure technologists often miss.

The doom scenario

The permanent underclass theory goes like this: Sam Altman has floated a $2,000 per month plan and eventually a $20,000 per month plan — what he calls a PhD in your pocket. If your competitors are using the $20,000 plan and you’re still skeptical about the free version, they might accumulate an insurmountable advantage. Then the $200,000 plan arrives, and you can’t even afford the entry fee. The gap compounds. The people who adopted early get richer, and everyone else gets locked out.

Taylor acknowledges the fear is rational. The spending disparity is already real. But he doesn’t think the doom loop will play out, and his reasoning is economic, not optimistic.

Three reasons the gap collapses

First, open-source models and price compression. AI model costs are dropping 90% year over year. A capability that costs $20,000 today will cost $2,000 next year and $200 the year after. The value of any specific intelligence tier collapses faster than anyone can build a moat around it.

Second, the physical world resists software. Every job that requires a human body — construction, healthcare, service work, anything involving physical presence — becomes more valuable as software-driven productivity inflates wages at the top. The economic ripple effect lifts physical labor, not replaces it.

Third, comparative advantage. If a $20,000 model exists, its owners won’t deploy it on tasks that a human can do cheaply. They’ll run it on frontier science, government applications, Fortune 500 problems. “It’s actually going to be cheaper for a company to hire me to do some data science,” Taylor explains, “because why would they want to retask that model away from foundational science?”

The punchline

“For how much it costs to feed us, we are AGI,” Taylor says. “You get AGI for what’s minimum wage.”

The argument isn’t sentimental. It’s a straightforward economic calculation: human intelligence is general-purpose, self-maintaining, and absurdly cheap relative to its capabilities. A frontier AI model costs thousands per month. A human costs food and shelter and produces general intelligence, social reasoning, physical labor, creativity, and unpredictable novelty.

Taylor pairs this with a more philosophical observation. In London’s St. Pancras train station, there’s a public piano. You could put a self-playing piano there — a solved problem even before AI — and program it with Bach and Mozart. Nobody would stop. But when a person sits down and plays, even badly, crowds gather. “Nobody stops for a self-playing piano,” he says. That gap between what’s technically possible and what humans find meaningful is, in his view, permanent.

How to position yourself

Taylor’s practical advice is counterintuitive: get off the laptop. Talk to people. Go to parties. Socialize. Build relationships. The things that no amount of AI spending can replicate — genuine human connection, trust built in person, content that matters because a real person made it — are the assets that appreciate as everything else gets automated.

He also recommends owning assets where possible (equity, property, stocks) and treating the current moment as an opportunity to develop skills that compound rather than skills that get automated. The people who will struggle aren’t the ones without AI access. They’re the ones who stopped building human relationships because they thought AI would handle everything.

FAQ

What is the permanent underclass theory in AI?

The theory posits that people who adopt AI early and spend more on premium tiers will accumulate compounding advantages, while late adopters fall permanently behind. With some individuals spending $4,600 monthly on 22 AI accounts and plans up to $20,000 per month announced, the concern is that an intelligence gap becomes an economic gap that can’t be closed.

Will AI create a permanent economic divide between early and late adopters?

Likely not. Three forces counteract the divide: open-source models and 90% annual price drops compress the cost of AI capability rapidly; physical-world jobs resist software automation and gain value as top-end wages inflate; and comparative advantage means frontier AI gets deployed on frontier problems, not on tasks humans can do cheaply.

Why does Mike Taylor say humans are AGI for minimum wage?

Human intelligence is general-purpose, self-maintaining, requires only food and shelter, and produces reasoning, creativity, social skills, and physical labor simultaneously. At minimum wage cost, a human delivers capabilities that would cost thousands per month to approximate with AI models — making human intelligence one of the most cost-effective intelligence sources available.

How should you prepare your career for an AI-dominated economy?

Force yourself to use AI for every task, even when it’s painful. Run blind tests comparing AI output to your own work to calibrate your actual assessment. Make space for play and experimentation. Own assets where possible. Most counterintuitively: invest in human relationships, physical presence, and social skills — these appreciate as AI commoditizes digital output.

Why are senior professionals at higher risk of falling behind on AI adoption?

Experienced professionals compare AI output against their own high standards and dismiss it as not good enough. Junior professionals, whose own output is already below AI quality, adopt immediately. By the time AI meets the senior professional’s bar, early adopters have years of workflow optimization. The expertise that was an advantage becomes a barrier to adaptation.

Will AI replace knowledge workers entirely?

Comparative advantage economics suggests otherwise. Even if AI surpasses humans at every individual task, deploying a $20,000 model on tasks a $15/hour human can do is economically irrational. Frontier AI will be directed at frontier problems. Human workers retain value for tasks where the cost-to-capability ratio favors biological intelligence, and physical-world work becomes more valuable as digital work gets cheaper.

Why does nobody stop for a self-playing piano?

In London’s St. Pancras station, a public piano draws crowds whenever a human sits down to play — even badly. A self-playing piano programmed with the greatest compositions in history would attract no one. The gap illustrates that humans value human effort and presence independently of quality or capability, a distinction that AI-generated content cannot bridge.

What assets should you build in an AI economy?

Equity in companies, property, financial assets where possible, but also human capital that AI can’t replicate: trust relationships, in-person networks, domain expertise with a human face attached. Content and skills that matter because a real person created them will appreciate as AI-generated alternatives become abundant and indistinguishable.