It’s Friday, and I’ve chased an article about the future of work down a rabbit hole. But here’s the gist, prompted by comments from Anthropic chief Dario Amodei last week that AI will create mass unemployment, and a response from Jim Amos, chief tech officer at the Human Intelligence Institute, that AI bosses selling AI ‘productivity’ (efficiency) tools should shut up about a social armageddon – when their motivation is to hype products and disrupt markets. “These billionaires aren’t working for the betterment of humanity, they’re just playing billionaire games,” he writes. 

Still, AI makers are talking very openly about how AI will disrupt labour markets, and the pace of adoption is rapid; the potential to replace cognitive labour at scale seems real. There should be a reset. The public sector, in concert with the private sector, has to get a grip of AI – to manage the mood, command the narrative, and model a framework for an AI-assisted economy. It can’t just be left to big-talking tech bros to manipulate yo-yoing markets that don’t know what’s going on (per the subtext in Amos’ commentary). We are approaching a structural break.

This isn’t ‘blue-collar’ automation to replace physical tasks; it is all-collar disruption to replace knowledge workflows, almost completely. If Amodei is right, however self-serving, then governments will be left to deal with mass unemployment. Clearly, I am not an economist or a politician (or a business person), but… it feels like we are on the brink of some BC/AD-style split, where the global economy might be usefully defined before and after this ‘new’ digital god called AI – like 2026 could be an economic ‘year zero’, after which everything changes. 

Perhaps 2026 is too soon; we are not yet seeing mass unemployment, and firms are cautious. But it could plausibly mark the start of a year-zero discontinuity window, when old narratives about reskilling and gradual adjustment begin to fail. Because by 2030, AI-driven efficiency gains will be measurable, labour compression will be visible, and policymakers will be forced to act. If it goes unchecked, the jobs burden will fall to the state. So here’s an idea (sketched on the back of an envelope, in need of proper research): tax AI efficiencies at 100 percent, both ways. 

Productivity and efficiency: different things, different impacts. Let AI productivity ring in corporate cash registers while taxing AI savings to fund a universal basic income. Take this year-zero line, 2026/2030, and measure the employment burden at every corporation, and keep it on their books as a combination of ongoing staff costs and new efficiency rents – even as headcounts shrink. Similarly, socialise profits from AI-efficiency products / companies while keeping private any growth (productivity) that expands new markets, particularly from startups and new ventures.

Point is: draw the line, year zero, and make any gains from labour-substituting AI projects subject to a levy to feed a basic income fund. Clearly, this is a sketch, possibly radical, probably naive. But it’s not a socialist anti-profit agenda; yes, it borrows one socialist intuition (socialised rents), but it operates inside a capitalist market system. It also asks more questions than it answers, notably about political feasibility: governments are aligned with capital in ways that make intervention difficult, and political will usually only aligns when systems stop delivering stability. 

But it feels like AI is about to see to that, anyway – 2026 to 2030, when pressure becomes visible in terms of labour disruption and rising inequality. But maybe AI is something society can plan for, somehow – rather than hoping markets will absorb the shock.