A ‘Golden Age’ for Whom?
AI ‘Golden Age’ or Power Grab? What do the latest interventions by Bezos, Khosla and Trump Actually Reveal
The last two weeks have seen a flurry of news stories that caught our attention, and trying to understand how they stack up against each other sparked an idea for our latest Substack
In the first article[i] (“Jeff Bezos says AI will bring ‘golden ages’ not mass job losses”), Jeff Bezos, the Amazon founder, dismissed predictions of mass job destruction and argued that AI will help usher in “multiple golden ages”, sitting alongside space, biotech and advanced manufacturing as part of a much larger wave of invention. It will not impoverish humanity. It will expand what humanity can do.
A few pages away (on the same day), the tone was rather different[ii] (“We will need a new tax code for the wealth AI creates”). Vinod Khosla, the venture capitalist and co-founder of Sun Microsystems, argued that AI will do 80 per cent of the economically valuable work humans currently do in 80 per cent of jobs. His conclusion was not that we should slow AI down; Khosla wants acceleration. But he also thinks the economic settlement around work, tax and ownership will need rebuilding, because AI may produce substantial underemployment and shift the rewards of growth further from labor to capital.
A few days later, Khosla appeared in a less policy-wonkish mode. Responding to Stanford students who walked out during a commencement speech by Sundar Pichai, chief executive of Google and Alphabet, he called the protesters “biased, idiotic, shortsighted, and very selfish”[iii]. Their protest focused on Google and Amazon’s cloud and AI contracts, and on wider concerns about surveillance and state power. Khosla’s response: the students were ignoring what he sees as the greatest opportunity for equality in human history, AI’s capacity to help the bottom three billion people on the planet.
And if that wasn’t confusing enough, on 12 June, the Trump administration issued an export control directive suspending all access by foreign nationals to Anthropic’s most advanced AI models, citing national security concerns[iv]. The signal to the rest of the world was hard to miss. AI is no longer just a commercial technology. It is becoming sovereign infrastructure.
So, in the space of a few days, we had four competing stories about artificial intelligence. Bezos gave us technological abundance; Khosla gave us economic disruption and redistribution; the Stanford protesters gave us power, accountability and harm; and Trump gave us AI as national leverage.
The temptation is to decide which one is right. But the uncomfortable possibility is that all four contain a truth. That is what makes the AI debate so difficult: it keeps collapsing into binaries, golden age or jobs apocalypse, utopia or dystopia. Yet the real world rarely offers such neat choices. Technologies can create wealth and destroy livelihoods; widen access and concentrate power; help one group while harming another.
Bezos’s argument is familiar, and not without force. Much of human progress has come from invention. Each wave changed the structure of work, but also expanded what societies could produce, imagine and build. AI is another general-purpose technology, making the impossible feasible. If Prometheus, Bezos’s new AI venture, can genuinely shorten engineering and manufacturing cycles, the benefits could be enormous. The problem with the golden age story is that it is incomplete.
Aggregate progress does not tell us who benefits, who pays, who adapts and who is pushed aside. The industrial revolution created huge wealth, but also grim factories and broken communities; the internet opened extraordinary access to information, but also gave us platform monopolies and surveillance capitalism. A technology can be historically beneficial and locally brutal at the same time.
This is where Khosla’s argument is more interesting than simple techno-optimism. He is not saying AI will fail; he is saying it may succeed so dramatically that the existing economic settlement no longer works. If AI and robotics reduce the demand for labor at a given level of GDP, the old bargain between work, income, tax and, what we might call, dignity begins to fray. In a world where capital owns the models, the compute and the robotics, the returns may flow upwards unless society deliberately designs a different mechanism.
We have spent decades telling ourselves that if people are displaced by technology, new jobs will arrive eventually. Often they have. But “eventually” can cover a lot of dislocation and pain — little comfort to the person whose skills are devalued at 48, or the graduate discovering the first rung of the professional ladder has been removed.
Khosla’s proposed answer is tax reform, near-free basic services and, eventually, broader public ownership of AI-generated wealth. The underlying question is hard to avoid: if AI creates vast value while requiring less human labor, how should that value be shared?
The Fortune article shows the harder edge of this worldview. Once AI is framed as the greatest equalizing force in human history, opposition starts to look morally suspect. If you believe AI will bring healthcare, tutoring and economic opportunity to billions, protests against the companies building it may appear parochial or self-indulgent.
But the students are asking a different question. They are not protesting AI as an abstract technology; they are protesting the institutional systems through which it is deployed — whether those systems become entangled with surveillance and state violence, who governs the technology, and who is harmed when powerful systems are sold into morally contested contexts.
Technology arrives in the world through contracts, incentives, institutions and business models, not as a pure expression of human potential. The same AI capability that helps a doctor diagnose illness might help a state monitor dissidents. The disagreement between Khosla and the students is not really about whether AI is powerful — both sides know it is. It is about whether the promise of future abundance should soften our scrutiny of present deployment.
The Trump administration’s reported action over Anthropic adds another wrinkle. If access to frontier AI models can be restricted by Washington on national security grounds, AI is not simply a technology market; it is part of the machinery of geopolitical power. That may be justified; frontier models are dual-use. But it is also naïve to miss the warning signal. If the most powerful AI systems are built by a handful of companies in one country, and access can be switched off by that government, the promise of global abundance comes with a geopolitical brake. Countries, companies and citizens outside the United States are not simply customers. They are dependents. Equalizing technologies do not equalize by magic; they equalize through access, affordability, governance and trust.
This is not an argument against AI, simply that its politics cannot be wished away by calling it progress. Once a technology becomes economically essential, socially embedded and strategically sensitive, it becomes infrastructure, and infrastructure always raises questions of power: who owns it, who governs it, and who has the right to challenge or refuse it?
The golden age narrative invites us to look past the messy present towards the shining horizon. Yes, there will be disruption. Yes, some institutions will behave badly. But history is moving forward; the long arc bends towards abundance. The arc, though, is shaped by people, institutions and incentives.
The protesters’ position also has a weakness. Walking out can be a powerful symbolic act, but it does not answer the question of how AI should be designed, audited, procured, constrained or used. If every engagement with a powerful technology company is treated as complicity, the space for practical reform narrows. We still need people who can ask better questions and spot second-order consequences.
We talk about jobs. We talk about tax. We talk about regulation, existential risk and national security. But we talk far less about the human capability needed to navigate the world being created.
Not the capability to use AI in the narrow sense — that conversation is already everywhere, with people trained to prompt, automate and build agents. The deeper capability is judgement under conditions of ambiguity. Can we hold several ideas in mind at once? AI may widen access to expertise for billions, while its infrastructure intensifies surveillance and coercion.
The great risk is that we train people to be efficient AI users without developing them as capable human judges: workers who can move faster, but not necessarily think better; organizations that can generate more outputs, but struggle to decide what matters, what is true, what is fair, and what should not be done.
That is why the future of human capital cannot be reduced to reskilling for tools — it requires a much more serious renewal of the human side of the human-AI equation.
In the AI-shaped workplace, the scarce human capabilities are likely to be the ones that sit above and around the machine: sensemaking, critical thinking, ethical reasoning, creativity and the ability to interrogate outputs rather than merely accept them — the ability to ask not only “can we?” but “should we?”, “who benefits?” and “what would make this decision wiser?”
This is where the binary debate fails us. Bezos is right to insist that invention matters. Khosla is right to insist that the economic settlement may need to change. The Stanford protesters are right to insist that technology cannot be separated from power. The Trump administration’s move reminds us that frontier AI cannot be separated from sovereignty.
The question, then, is not whether AI will bring a golden age. The better question is: golden for whom, governed by whom, accessible to whom, and at what cost? And what human capital should we develop to maximize our chances of staying relevant?
This is the work PolymathMind does — helping senior leaders close the gap between AI adoption and the human operating model around it. If your organization is wrestling with what to keep meaningfully human, how to protect the routes through which expertise is built, or how to make accountability real rather than decorative, those are conversations we’d welcome.
[i] https://www.ft.com/content/45258f4a-5fb5-4f7a-bd29-5dea92502638?syn-25a6b1a6=1
[ii] https://www.ft.com/content/b277360e-bf23-4366-afd7-acab940f66b7?syn-25a6b1a6=1
[iii] https://fortune.com/2026/06/15/vinod-khosla-stanford-protest-google-sundar-pichai-walkout/
[iv] https://edition.cnn.com/2026/06/13/business/anthropic-mythos-model-national-security