Do We Really Want a No-Hands Job From Silicon Valley? Who Holds the Power in the Age of AGI

“Anyone that’s properly using AI now knows that you tell it what you want, it gives you a plan, carries out the work, and you judge and tweak. You’re not a passive victim — you’re an active user with outcomes in mind.” — Keith Teare

Do we really want a no-hands job from Silicon Valley? That Was the Week newsletter publisher Keith Teare — who thinks all tech innovation results in human progress — thinks we do. No hands, no problem, Keith says. But I’m not sure. Especially given the powers-that-be giving us that no-hands job.

Keith welcomes the end of what he calls the “typed” and “touched” computing era — keyboards, mice, touchscreens, and all the manifold ways we have used our hands to interact with computers since the 1980s. That’s the outcome, he predicts, of the race to AGI. So far so good. But what happens if our no-hands AI future is controlled by Google, Microsoft, Amazon, and Facebook? This week these four behemoths committed 00 billion to AI infrastructure investment in 2026 alone — 2 percent of all US GDP. These companies are racing to build (and own) the foundational mechanics of AGI.

That’s always how it’s been, Keith says, embracing our no-hands future. I’m less open-armed. What happens if we want our hands to fend off AGI? No, I’m not so keen on a no-hands job from Silicon Valley. Especially one couched in the altruism of human progress.

Five Takeaways

•       The End of the Hand-Driven Computing Era: Andrej Karpathy’s observation at Sequoia’s AI Ascent: he no longer uses his hands to do his work. He speaks to the computer; the computer acts; he judges and refines. The keyboard, the mouse, the touchscreen — all the hand-driven interfaces that have defined computing since the 1980s are entering their twilight. Karpathy calls it “software 3.0”. Keith, two years ago, wrote an editorial called “eyes, hands, ears, and mouth” about the inclusion of other human attributes beyond hands. That prediction has arrived.

•       $700 Billion: The CapEx Explosion: A post by @Signal framed the week’s numbers: $700 billion in AI infrastructure spending in 2026, equivalent to 2 percent of all US GDP. This kind of spending, the post observes, usually happens via governments or wars. This time, it’s four private companies — Microsoft, Amazon, Google, and Meta — racing to build the foundational mechanics of AGI. Meta was punished by Wall Street for overspending; Google was rewarded because its numbers were strong enough to justify it. The same bet, two different verdicts, depending on your quarterly earnings.

•       Was the Internet Privately Built? The ARPANET Argument: Keith’s claim: innovation waves have always been privately financed. The railways, the telephone, the electricity grid, the commercial internet. Andrew’s counter: ARPANET was a massive government investment that created the protocols on which the internet runs. Keith’s response: ARPANET was a university bulletin board that created the precedent, not the infrastructure. Andrew’s response: that’s not exactly what ARPANET was. They agree that government research matters. They disagree on how much credit it deserves for what became the commercial internet.

•       The Revenge of the Idea Guy: Sam Altman’s line of the week. In the past, an idea person came up with a concept and then needed expensive engineers to build it. Many ideas never saw the light of day because the engineering cost was prohibitive. Now, anyone can speak an idea into existence. AI builds the plan, executes the work, and you judge and refine. That changes the economics of creativity, advertising, software development, and anything else that used to require specialist execution. The specialist is not dead — but specialists