The Real Question

Picture the economy as one big pie that everyone in a country shares. Right now, AI is doing something remarkable to that pie: it's making it bigger, faster than almost anything in living memory has. Lawyers can draft in minutes what used to take days. Factories can plan output with less waste. Doctors can catch things earlier. All of that is real, and all of it means the pie, the total amount of stuff and services the economy produces, is going to keep growing.

Here is the only question that actually matters: when the pie gets bigger, who gets a bigger slice?

If the extra slice goes almost entirely to the small number of people and companies who own the AI, while the rest of the pie doesn't grow proportionally to how many mouths need feeding, that's a problem, not because the pie shrank, but because most people's slice didn't grow with it, and might even shrink if their job disappeared along the way. If the extra slice gets shared out broadly, through higher wages, new kinds of jobs, dividends, or direct payments, then a bigger pie is exactly as good as it sounds.

Notice what that question is mostly not about. At the pace AI is actually improving right now, it's not about how smart the AI gets. A superintelligent AI and a mediocre one can each end with the pie shared fairly or shared badly, in today's world, the thing that decides it is who owns the slicing knife, not how big the oven is. (There's one edge case, an AI so explosively productive that sheer growth outruns any sharing problem on its own, but nothing on the horizon looks like that yet, so treat it as a footnote, not the plan.)

Why I'm Calling This STHIRTA

For fifteen years, consultants and executives have described the world with one word: VUCA, a term the US military coined decades ago for Volatility, Uncertainty, Complexity, and Ambiguity. Everything felt faster, less predictable, harder to plan around. AI forces a choice between two futures: a louder version of the same VUCA, or something this piece calls STHIRTA (स्थिरता), the Hindi word for stability, a state of equilibrium. The name comes from the Fifth Industrial Revolution: read the "5" as it looks, close to the letter S, and "5THIR" becomes "STHIR", स्थिर, Hindi for stable.

AI is the technology that will answer whether we get STHIRTA or more VUCA, one way or the other, for our generation. This piece is about what "answering it well" actually requires.

This framing isn't new to this piece: I first argued a version of it at The Fifth Estate Summit in Delhi in December 2017, in a concept paper on the "5th Industrial Revolution," and in a talk the following year titled "From VUCA to 5THIR-TA."

Try It: Slice the Pie Yourself

Two things decide whether we land closer to STHIRTA or closer to VUCA: how much of AI's extra wealth gets shared back with people who didn't own the AI, and how fast jobs are changing while that sharing catches up (or doesn't). Move the sliders.

Your Washing Machine Still Needs You

Here's something almost every household already knows without having a name for it. A washing machine automates the washing. A dishwasher automates the washing-up. A robot vacuum automates the sweeping. But none of them sort your laundry by color, load the machine, move wet clothes to the dryer, fold them, put them away in the right drawer, or navigate around your kid's toys on the floor. All of that, the physical, unpredictable, in-between part of the job, is still entirely on you.

There's a real, well-known reason for this gap, called Moravec's paradox, after roboticist Hans Moravec, who described it in his 1988 book Mind Children (with related observations from AI researchers Rodney Brooks and Marvin Minsky around the same time). The paradox is this: tasks that feel hard to us, chess, advanced math, legal reasoning, are actually cheap for a computer. Tasks that feel effortless to us, picking up an object, walking across a cluttered room, telling a clean shirt from a dirty one by feel, are extremely expensive for a machine, because those skills were built by a billion years of evolution, while abstract reasoning is a very recent add-on. This is why "AI" and "a robot that can do your laundry end to end" are really two different technologies moving at two very different speeds, one being called physical AI to distinguish it from the language-and-reasoning AI you've been reading headlines about.

Physical AI is arriving, just much more slowly, and the prices as of 2026 are genuinely real, not hypothetical:

Note that last line carefully: Tesla's own stated target price is $20,000-$30,000, but independent analyst estimates of what it actually costs to build one today run $50,000-$100,000 or more, with the first commercial units expected to sell for $100,000-$150,000, well above the eventual target. The gap between "eventually cheap" and "cheap now" is exactly where the real economics live.

One more thing a simple sticker-price-divided-by-months calculation leaves out entirely: financing and upkeep. In a market like India, where financing costs run well above what a developed-market buyer pays, and where a first-generation robot's maintenance and electricity draw are real, ongoing costs, not one-time ones, the true monthly cost of owning a robot runs meaningfully above the naive amortized figure. A human hire is also, unlike a robot, an operating expense you can simply stop paying next month if your own income disappears; the robot is sunk capital either way. The comparisons below layer on an illustrative total-cost-of-ownership premium that scales with the local-wage slider, higher toward the India end to reflect steeper local financing costs, import duties, and thinner repair networks, lower toward the US end, which is one more reason human help stays the cheaper option in India for longer than the sticker prices alone suggest.

Will a Robot Replace Your Maid?

This is where household economics gets genuinely interesting, and it's not the same answer everywhere. Economist John Hicks argued back in 1932 that new technology gets adopted fastest wherever it replaces the most expensive input, which is why industrial robots swept through high-wage manufacturing countries, Japan, Germany, South Korea, decades before they showed up in low-wage ones.

The same logic applies to your living room. In a typical middle-class Bangalore household, a part-time maid and part-time cook together often cost somewhere around ₹12,000-15,000 a month. Compare that to $449 a month (roughly ₹37,000) for the Weave Isaac 1, or the monthly-equivalent cost of a $16,000 Unitree G1. In India, the human option is currently far cheaper, and comes with something no robot has yet: the ability to climb stairs, adapt to a cluttered, unfamiliar layout, and improvise. In a country like the United States, where full-time domestic help can run $3,000 a month or more, that same robot subscription is an obvious, immediate bargain. So the honest prediction is: this "supply push" of humanoid robots lands first and hardest in high-wage countries, and much more slowly in India, not because Indian households don't want help, but because the math doesn't favor the robot yet.

Here's the part that connects everything back to the pie. Say you earn the equivalent of $2,000 a month at your job. An hour spent doing your own laundry is an hour not spent earning at that job, so paying someone $2 to do it instead is an easy call. Now say you lose that job. The hour of laundry doesn't get any faster to do, but the job you'd otherwise have been working no longer exists, so there's nothing being given up by doing it yourself anymore. Economists call what you're weighing here "opportunity cost", what you give up by choosing one thing over another, and economist Gary Becker showed, in a 1965 paper that's still taught today, that this exact trade-off is what decides whether a household pays for help (human or, now, robotic) or does the task itself. When someone has a good job, hiring help wins. When that job disappears, to AI, in this case, the math flips overnight: doing it yourself becomes cheaper than paying anyone, human or machine. That's not a mood or a belt-tightening decision, it's the rational, predictable result of a sixty-year-old economic model meeting a sudden loss of income.

Try It: Does a Robot Make Sense Where You Live?

Set your local monthly cost for hired help and see how it stacks up against real robot prices.

We've Done This Before, and It Wasn't Easy

Economic historian Robert Allen coined a term, "Engels' Pause," named after Friedrich Engels, who documented it firsthand, for a real and well-studied period of the British Industrial Revolution, roughly the 1790s to the 1840s. During those four-plus decades, output per worker rose by about 46%, and profits roughly doubled as a share of national income, while ordinary workers' real wages grew by only around 12%, essentially flat for an entire working generation. The long-run outcome was genuinely good, higher wages, the modern middle class, but an entire generation lived and worked through the pause before that arrived. That's the honest historical answer to "won't new jobs eventually appear": probably yes, eventually, but "eventually" has, at least once before, meant a working lifetime.

On the more hopeful side, there's already a real, working example of the kind of sharing this piece keeps calling for, not a theory, an actual functioning program. Alaska's state constitution created the Alaska Permanent Fund in 1976, and since 1982, it has paid every single resident of the state an annual dividend from the state's oil revenue, no application, no means test, just being a resident. It's been running for over four decades. If someone tells you an "AI dividend" is a naive fantasy, Alaska is the answer: a version of exactly that idea has already been running, successfully, for longer than most people reading this have been alive.

So What Do We Actually Do?

If the whole argument boils down to "share the bigger pie," it's fair to ask what that looks like in practice. It's not one thing, it's a menu, and different countries will likely reach for different combinations:

None of these happen automatically. Every one of them requires a deliberate choice, usually a political fight, for the same reason given above: markets don't sort this out on their own.

Even If We Get This Right, One Thing Won't Be Solved

Suppose we do all of it well. The pie gets shared, the dividend gets paid, the transition doesn't take a whole generation. Does that actually make people content?

In 1930, John Maynard Keynes, the most influential economist of his century, wrote an essay called "Economic Possibilities for our Grandchildren," predicting that within about a hundred years, productivity gains would be so large that people would work "three-hour shifts, or a fifteen-hour week," spending the rest of their time on things that actually mattered to them. He was right about the productivity. He was wrong about the 15-hour week, and the reason why is genuinely useful to understand.

In 1949, economist James Duesenberry showed that people don't judge their own well-being by some absolute standard, they judge it against their neighbors and their own recent past. That's the relative income hypothesis, and it means rising abundance doesn't end status competition, it just moves the goalposts. Even in a world where AI's gains are shared fairly and nobody is poor by any historical standard, someone whose neighbor has a slightly better version of everything will still feel behind. That's not a flaw in the sharing, it's just a limit on what sharing wealth can fix. Economic policy can guarantee everyone has enough. It can't guarantee everyone feels like they have enough.

If You Lead a School, College, or University

Here's why an education-strategy consulting firm is the one writing this. The washing-machine economics earlier in this piece aren't abstract: the same disposable income that decides whether a Bangalore household hires a maid and a cook is, for a huge share of India's urban middle class, generated by the entry-level IT-BPM ladder this piece keeps coming back to. That income is also what pays premium K-12 and university fees. If AI compresses that ladder faster than the sharing mechanisms this piece describes catch up, institutions that have quietly priced themselves on "we prepare you for a BPO or coding career" are pricing themselves on the exact segment of the economy most exposed to the fork. The curricula that survive this shift are the ones that pivot toward what AI is still bad at, complex judgment, physical adaptability, human trust, well before the ladder actually gives way, not after. That's the specific, practical question behind everything above, and it's worth a second opinion before it shows up in your admissions numbers.

So, STHIRTA or VUCA?

Before we finish, it's worth being honest about which parts of this are settled and which parts are still a bet. Moravec's paradox, Baumol's cost disease, the Prisoner's Dilemma, Engels' Pause, the Alaska Permanent Fund, these are documented history and tested economics, not predictions. What's genuinely uncertain is how this specific transition plays out: whether new jobs appear fast enough this time, whether the sharing mechanisms get built before the strain sets in, whether AI-specific job losses move faster than every past technological shift. Anyone who tells you they know that part for certain, in either direction, is selling you a confident story, not an honest one.

What is within anyone's control is the choice this piece keeps returning to: build the sharing mechanisms deliberately, or don't, and live with whichever version of the future that choice produces. From VUCA to STHIRTA was never going to be automatic. It never is.

If you want the full formal version of this argument, with the actual equations, the named economic model, and an interactive calculator you can stress-test with your own numbers, read the companion piece: The Ownership Fork.

Sources

1. Merrill Flood & Melvin Dresher (1950), RAND Corporation; the Prisoner's Dilemma was later named and formalized by Albert W. Tucker.
2. Mancur Olson (1965), The Logic of Collective Action.
3. Irving Fisher (1933), "The Debt-Deflation Theory of Great Depressions," Econometrica.
4. William Stanley Jevons (1865), The Coal Question.
5. Hans Moravec (1988), Mind Children: The Future of Robot and Human Intelligence; related observations by Rodney Brooks and Marvin Minsky.
6. John Hicks (1932), The Theory of Wages (induced innovation).
7. Gary Becker (1965), "A Theory of the Allocation of Time," The Economic Journal.
8. David Ricardo (1817), On the Principles of Political Economy and Taxation (comparative advantage).
9. William Baumol & William Bowen (1965/1966), "On the Performing Arts," American Economic Review, and Performing Arts: The Economic Dilemma.
10. Robert C. Allen (2009), "Engels' Pause," Explorations in Economic History.
11. Alaska Permanent Fund, established 1976; dividend program since 1982.
12. John Maynard Keynes (1930), "Economic Possibilities for our Grandchildren."
13. James Duesenberry (1949), Income, Saving and the Theory of Consumer Behavior.
14. Daniel Ellsberg (1961), "Risk, Ambiguity, and the Savage Axioms," Quarterly Journal of Economics.
15. Weave Robotics, Unitree Robotics, and Tesla product pricing as publicly listed, 2026.

Prices, wage estimates, and workforce figures are illustrative and current as of publication; verify against the original source before citing elsewhere. The India domestic-worker figure is a known undercount per government surveys; independent estimates vary considerably.