Writing at roughly 6,000 words on Gates Notes, Bill Gates called for institutions at both national and international level, for job categories he labels “Human Reserved”, and for taxes falling on robots and on AI tokens, all to soften a transition he described as potentially among the most disruptive humanity has been through. About two weeks later, on September 8, the outlet X.PIN ran a rebuttal by journalist CT Zhao arguing that Gates’s framework targets a future that has not arrived while leaving the present nearly unregulated.
The disagreement is not really about whether AI will eventually reshape work. Gates has not walked that claim back, and Zhao does not dispute it either. What Zhao contests is Gates’s implied timeline: that the disruption is already so far along that the right response is a new global tax-and-institution architecture rather than scrutiny of the companies making layoff decisions right now.
Zhao’s central claim is that current layoffs reflect management choices, not machine capability outrunning workers. For evidence the piece reaches for Klarna, whose 2024 claim was that its assistant fielded 2.3 million conversations inside one month and might add $40 million to annual profit. By 2025, according to Zhao, chief executive Sebastian Siemiatkowski conceded the company had cut too aggressively and moved to bring back remote human agents, with Klarna’s own regulatory filings now describing a support model that blends software with people. Zhao also points to Meta’s reported “Project OT,” an internal push to rebuild teams around AI tools, where code commits reportedly climbed 220 percent even as technical and security incidents rose 40 percent, and where a wider layoff round originally slated for November was reportedly put on hold.
Both examples are single-company anecdotes, not labor-market data. The piece does not claim to have counted total AI-linked layoffs or measured them against hiring elsewhere in the economy.
On the more alarming end of the debate, Zhao notes that Anthropic chief executive Dario Amodei has forecast that half of all entry-level office jobs could disappear within a handful of years because of AI, a claim Nvidia chief executive Jensen Huang has dismissed as reflecting a “God complex” among AI executives. Zhao sides with skepticism here, arguing that a lab predicting its own product’s impact is not a disinterested source.
Zhao’s most concrete evidence concerns biological risk, an area where Gates’s essay reportedly voiced concern about AI-enabled bioterrorism. Zhao cites a Stanford University and Arc Institute project in which genome language models generated candidate bacteriophages capable of killing E. coli strains: out of roughly 300 lab-synthesized designs, only 16 proved viable, and the reference genome the team worked from ran to 5,386 nucleotides, a fraction of the roughly 30,000 in SARS-CoV-2. Zhao treats this as evidence that turning a model output into a functioning pathogen still demands laboratories, funding and specialist skill that most actors lack, a fair reading of a genuinely narrow proof-of-concept rather than a demonstrated bioweapon shortcut.
Turning to money, Zhao reports a $15 billion pre-IPO credit line being finalised, plus a separate arrangement of about $35 billion in private credit earmarked for buying chips, part of it backstopped by Broadcom. He also points at Leading the Future, an industry super PAC reportedly holding over $125 million for the 2026 cycle from backers including Greg Brockman and Andreessen Horowitz, against Anthropic’s $20 million to Public First Action, which the company says funds public education rather than campaigning.
Where Gates proposes taxes on future AI tokens and robots plus new multinational bodies to administer them, Zhao’s counterargument is that regulators should instead target the hyperscalers, data-center operators and financiers already identifiable today, through grid and water fees, tax-abatement reviews and labor-transition funds. Neither position is settled policy, and both remain contested: taxing a technology that has not scaled to Gates’s imagined endpoint could tax the wrong thing, while regulating today’s builders does nothing about tokens or robots that do not yet dominate the economy Gates is describing.
Much of this fight is two people answering different questions with the same word. Gates is describing a multi-decade equilibrium in which AI capability eventually does displace large categories of work; Zhao is describing the next few quarters of corporate hiring decisions. A claim about 2030s labor markets and a claim about this year’s earnings calls can both be accurate without contradicting each other, which is why the essay and its rebuttal talk past one another as often as they engage.
For operators, the practical takeaway is to treat any AI-driven headcount claim, whether from a vendor or a rival’s earnings call, as a management decision to verify rather than a capability benchmark to accept, and to watch which regulatory approach, token taxes or current-company oversight, gains traction first in your jurisdiction over the next two quarters.
X.PIN, in a piece bylined X.PIN and CT Zhao published September 8, 2026.
