Cory Doctorow used a July 13 essay on Pluralistic to press a basic question at AI companies: if their systems can replace doctors, accountants, teachers and other workers, why are they selling software to the employers of those workers instead of taking the work themselves?
The argument targets the gap between the industry’s most expansive claims and its actual commercial behavior. Doctorow writes that AI promoters describe systems on a path toward general intelligence, while still asking customers to pay for chatbot services that would, in theory, cut labor costs for hospitals, tax-prep firms, schools and other institutions.
His point is not a benchmark dispute or a model-card nitpick. It is an economic test. If an AI company already owns a tool that can perform a paid job more cheaply than a human worker, Doctorow argues, the company should be able to sell the finished service directly and keep the margin, rather than sharing the proceeds with an incumbent business that buys the tool.
Doctorow frames the question through an old “picks and shovels” story: during a gold rush, selling supplies to miners can be safer than mining. He pushes the metaphor one layer further, arguing that the richest position is often wholesaling supplies to the retailers who sell to miners. In his telling, much of the modern economy rewards distance from the actual work.
He ties that to Douglas Rushkoff’s phrase “go meta,” which Doctorow describes as a business culture that prizes abstraction over doing the thing itself. His examples move from driving a taxi, to owning a medallion, to operating a ride-hailing company, to investing in that company, to trading options on the investment. Each step sits further from the service being delivered.
Doctorow applies the same critique to AI vendors. If a chatbot can do a doctor’s job, he asks why the vendor would sell it to a hospital rather than open a hospital. If it can do tax preparation, why sell it to a tax-prep company. If it can teach children, why sell it to a school district.
The essay also takes aim at AI rhetoric around scientific progress. Doctorow says figures he identifies as Altman and Amodei have suggested, in broad terms, that advanced AI systems could become extraordinarily powerful. He also alleges that DOGE staffers dismissed National Institutes of Health cancer research projects on the grounds that general AI would soon solve cancer. Those claims are presented by Doctorow as examples of the industry’s inflated sales posture, not as confirmed evidence that such systems can perform those tasks.
Doctorow compares the AI sales pitch with get-rich-quick schemes and stock-picking claims: if the seller has found a reliable way to generate outsized returns, he asks, why sell access instead of using it. The same suspicion drives his AI critique. A vendor that really had cheap, competent automated labor would have reason to compete directly in the markets it says it can transform.
His conclusion is that the industry’s current structure looks less like proof of near-omnipotent software and more like another layer of the meta-economy: companies selling “doctorbots” to hospital managers, rather than proving the bots can run the clinic.
This story draws on original reporting from Pluralistic.