Why Did the King Run Out of Gold?

The pace of technological change, and the time society needs to adapt

Joo-Ho Lee

1. The Arithmetic of a Single Gold Coin

Once upon a time, a king invited a subject who had served him well to name his reward. The request sounded surprisingly modest: one gold coin today, two tomorrow, four the day after. Nothing more than twice the previous day’s payment, every day. The king smiled and agreed on the spot.

For the first ten days, there was hardly anything to worry about. The payment on day ten was 512 coins, and the total paid out so far was just 1,023. Small change for a royal treasury. By day twenty, however, that day’s payment alone came to 524,288 coins. The running total had passed one million. At eight grams per coin, that was about 8.4 metric tons of gold. Enough, perhaps, to make the treasurer take another look at the books.

But the real story unfolds over the next ten days. On day thirty, the king owes roughly 537 million coins in a single payment. The total for all thirty days is about 1.074 billion coins, weighing some 8,600 metric tons. That is more than the roughly 8,100 metric tons in the United States’ official gold reserves. By day thirty-five, the total weight exceeds 270,000 metric tons. Even all the gold humanity has ever mined would not be enough to keep the promise.

The point is not simply that the numbers get big. When the payment doubles every day, today’s payment exceeds everything paid out before it. The payment on day thirty is exactly one coin more than the combined payments of the previous twenty-nine days. A promise that initially seemed manageable becomes an entirely different problem.

I think of this story when I look at technological change today. Of course, technologies do not all double at a steady rate like those gold coins. Still, the fable raises a question: are we judging tomorrow’s changes only by the size of what we see today?

2. The Long, Slow Beginning

The oldest stone tools currently known were found at Lomekwi in Kenya and date back about 3.3 million years. Imagine compressing those 3.3 million years into a single twenty-four-hour day. At midnight, someone chips a stone to make a cutting edge.

Most of the day passes. Ways of shaping stone change; knowledge of how to use fire accumulates. There is change, but viewed from the present, the intervals between developments seem almost unimaginably long.

The beginnings of agriculture, around 12,000 years ago, arrive at about 11:54:46 p.m. on this clock. A little over five minutes remain before midnight. Early writing in Mesopotamia, around 5,200 years ago, appears at 11:57:44 p.m., just over two minutes before the day ends.

Then familiar milestones crowd into the final seconds. Gutenberg’s development of metal movable-type printing in Europe in the mid-fifteenth century comes about fifteen seconds before midnight. Watt’s patent for improvements to the steam engine in 1769 arrives about seven seconds before midnight; the Wright brothers’ first powered flight in 1903, about three seconds. Taking 2026 as our present, the public release of ChatGPT in late 2022 falls roughly a tenth of a second before midnight.

Much of what we consider everyday life fits into those last few seconds. Changes that once accumulated over countless generations now arrive one on top of another within a single lifetime. We have correspondingly less time to understand new tools and develop the customs and rules to live with them.

3. The Time Society Needs

New technologies have always unsettled society. Printing changed the routes by which knowledge spread; the automobile changed the routes by which people and goods moved. Neither arrived with a finished set of rules and ways of life to match.

As printing houses spread across Europe from the mid-fifteenth century onward, books and debates reached wider audiences. Printed material also played an important role in the Reformation. A change in the tools for making books helped broaden arguments about religious authority and political order. Adjusting to this new information environment was a long and sometimes turbulent process.

The changes surrounding the automobile are more concrete. Britain’s so-called Red Flag Act of 1865 required a person carrying a flag to walk ahead of mechanically propelled vehicles on the road. It was one response to the problem of fitting unfamiliar machines into existing roads. Britain introduced driving licenses in 1903 and compulsory driving tests in 1935. Traffic signals, insurance, seat belts, and other measures found their place at different times.

None of this was a peaceful process. Industrialization brought labor exploitation, accidents, and pollution, and some of the damage went unaddressed for a very long time. It would be hard to argue that societies in the past kept up with technology particularly well. But when a technology took years to build out physically and bring into widespread use, there was relatively more time to try out institutions and revise them.

Digital technology is changing those conditions. Delivering a new feature to people who are already connected does not require a new factory or a new road. While society is still looking for answers, the technology can move on.

4. The Later Days of the Gold-Coin Story

On December 17, 1903, the Wright brothers’ first flight covered about thirty-seven meters in twelve seconds. It would not have crossed a football field. Sixty-six years later, in 1969, humans set foot on the Moon. A child who heard the news of that first flight could have watched the Moon landing on television in old age. Flight and space travel had both arrived within a single lifetime.

The numbers in semiconductors tell an equally striking story. Intel’s 4004 microprocessor, released in 1971, contained around 2,300 transistors. NVIDIA’s Blackwell GPU, announced in 2024, contained 208 billion across two chip dies. That is about ninety million times as many. Transistor count, of course, is not the same as performance, nor is it a measure of progress across science and technology. It does, however, show how much the physical foundation of computing has changed.

Artificial intelligence developed on that foundation. In 2016, AlphaGo defeated the ninth-dan Go master Lee Sedol by four games to one. AlphaGo Zero, announced the following year, learned by playing against itself without studying records of human games. It defeated the version that had played Lee Sedol by one hundred games to zero.

Released in late 2022, ChatGPT brought this change out of the laboratory and into everyday life. Anyone could ask a question on the web and get an answer. The internet and smartphones had already laid the paths along which it could spread. Yet trying a service once is very different from changing how a school or a business works. Access can grow quickly; learning to use a tool well and deciding who is responsible for what take longer.

The way research is done has also begun to change. The AlphaFold database has made structure predictions available for more than 200 million proteins. A prediction is not experimental confirmation. Even so, researchers now have a vastly larger set of starting points for forming hypotheses and deciding which experiments to pursue.

This is what draws my attention. Technology has long helped us create the next technology. Now AI assists with parts of that process, including design, coding, and data analysis. That opens up the possibility of shortening the time it takes to develop what comes next. Experiments and validation still take time; energy and cost remain constraints. But if the process of making new technologies accelerates too, the next change may arrive before society has grown accustomed to the last one.

5. Struggling to Keep Our Balance

Law, morality, and culture move at a different pace from technology. Laws must reconcile competing interests, and customs take shape through repeated experience. That slowness is not necessarily a flaw. Part of what makes institutions dependable is that they do not change on a whim. The question is whether we can preserve that care while responding to change.

The European Union’s AI Act illustrates the difficulty. The European Commission proposed the legislation in April 2021. ChatGPT appeared in late 2022 while the proposal was still being debated, and the final law included provisions for general-purpose AI models. The law entered into force in August 2024 and is being applied in stages. Even as lawmakers were writing it, the technology and market they sought to regulate were changing.

Deepfakes offer a more painful example. People have been harmed by the spread of sexual images made using their faces without their consent. In the United States, the TAKE IT DOWN Act became law in May 2025, criminalizing the publication of certain nonconsensual intimate images and requiring platforms to provide removal procedures. Some existing laws already applied, but gaps remained in how adequately they addressed these new forms of harm. For victims, those gaps were far from an abstract problem of institutional design.

The questions of morality and culture are harder still. In a work made with AI’s help, where does human creativity begin and end? Is a student’s AI-written report cheating, or is it like using a calculator? Does recreating a deceased family member’s voice with AI bring comfort, or make grieving harder? The answers will vary with the context and with how the technology is used.

What about an apology written by AI? The words might help someone express genuine remorse, or give them a way to avoid responsibility. New tools also reshape our sense of what counts as sincerity and good manners.

There has always been a gap between technology and society. Now, before we have narrowed it, another change arrives. While we amend one law and get used to a new set of social expectations, technology is already posing the next question. We need to ask whether catching up after the fact is enough, and what we can do to prepare beforehand.

6. Doing the King’s Arithmetic Differently

The king’s mistake was to judge the whole promise by the payments of the first few days. Our intuition can lead us into the same trap. We assume that something manageable today will remain manageable tomorrow. Or we make the opposite mistake: we see rapid change in one area and assume everything will change at the same pace.

First, we need to look beyond the current scale of a technology to the rate at which it is changing. Rather than stopping at “This seems fine for now,” we should ask, “What happens if this pace continues for a few more rounds?” Instead of assuming that AI will remain bad at the things it does badly today, we need to consider what would change if it became good at them.

That does not mean betting everything on a single future. Some changes arrive sooner than expected; others hit a wall. Governments should prepare for several possible futures in employment, education, and social welfare. Individuals would do well not to depend on a single tool or skill. More important than getting a prediction exactly right is leaving ourselves room to respond when it turns out to be wrong.

Our institutions also need to be easier to revise. We can build in procedures from the outset for reviewing the benefits and harms of a policy after it takes effect, and for changing it when necessary. Standards for safety and rights must hold, while the means of meeting them should be able to change in light of new evidence. Rather than waiting for a perfect answer, we need ways to learn and improve as we go.

For individuals, the habit of learning becomes more important. We need to try new tools, judge how they might change our work, and resist taking plausible-looking results at face value. Children need to learn to ask questions, verify answers, and work with others, just as much as they need to learn the right answers.

Finally, amid all that is changing, we must not lose sight of what is worth preserving. New tools cannot take over the work of building trusting relationships, trying to distinguish right from wrong, or accepting responsibility for our choices. Knowing what a technology can do is different from deciding what we should do with it. That decision is still ours.

Epilogue: A Morning Around Day Twenty

The most frightening moment in the gold-coin story is not day thirty-five. It is that morning around day twenty when the treasurer first takes another look at the books. The gold paid out by then might still seem manageable. But to anyone who has calculated what the remaining days will cost, the promise already looks very different.

We cannot know exactly which day we are living through. Technology’s future has neither the fixed doubling rate of the gold coins nor a catastrophe written into the arithmetic. What is clear is that today’s pace alone cannot tell us what tomorrow will bring. We can choose whether merely to catch up after change arrives, or to begin changing our institutions and the way we learn before it does.

Rather than closing the ledger and living today as we lived yesterday, it is time to look at the next page together. If the king had worked through the numbers just once before smiling and saying yes, the story might have ended differently. We, too, still have time to calculate and choose.

Sources and further reading

World Gold Council · Harmand et al., Nature (2015) · UK road safety history · NVIDIA Blackwell · AlphaGo Zero · AlphaFold Database · EU AI Act · TAKE IT DOWN Act

Illustrations are AI-generated conceptual images. The clock is symbolic, not a precisely scaled timeline.