Fifty-Two Million Characters

Why the autonomous revolution needs comparative religion more than another ethics framework

In 1236, the government of Korea was living on an island.

Five years earlier, Mongol cavalry crossed the northern border, forcing King Gojong’s court to leave the mainland capital for Ganghwa, a tidal island safe from the horsemen. On the mainland, the invaders burned everything they found, including the woodblocks of the first Korean Buddhist canon, which had taken seventy-six years to carve.

The court, cornered and losing a war it would go on to lose, ordered the whole thing cut again.

The project lasted sixteen years, from 1236 to 1251. Artisans on Ganghwa and the southern coast cut birch trees, soaked the logs in seawater for three years, then cut them into blocks, boiled them in brine, and dried them in the shade for another three years before carving. They made 81,258 blocks, carving both sides, for about fifty-two million characters. The work is so consistent that tradition says one person carved it all.

As war magic, it failed. Goryeo capitulated in 1259. Kim Chun had assassinated Choe Ui, the military strongman whose family sponsored the carving, the year before. The court came off the island under Mongol edict, and Korea spent the next century as a vassal.

The blocks survived. In 1398, they were moved to Haeinsa, a temple on Gayasan in the south. By 1488, monks had finished the four buildings that still house them. These buildings are especially interesting. They sit about 655 meters above sea level, positioned to catch dry mountain air and avoid damp valley winds. The north and south windows differ in size, turning the hall into a natural ventilation system. The packed earth floors, layered with charcoal, lime, salt, and sand, draw moisture from the air during the wet season and release it in the dry.

In 1970, the Korean government built a new storage facility for the blocks, using modern materials and climate control. Officials moved some test blocks inside, but mildew developed. They returned the blocks to the 1488 hall, where they remain today.

Once again, we are constructing something on a scale the original carvers could never have dreamed of, yet we speak of it as if it were nothing more than pipes and fittings.

The building is the belief.

The data center stands as the cathedral of our era, built with the offerings of people who may never know if their faith was justified.

The International Energy Agency estimates that global data centers will use about 415 terawatt-hours of electricity in 2024, about 1.5 percent of the world’s total. This is growing by about 12 percent each year and could reach 945 terawatt-hours by 2030. In the United States, data centers are expected to drive nearly half of all electricity demand growth through 2030, and by then, they may use more power than the aluminum, steel, cement, and chemical industries combined. Water use is harder to track, since most operators do not report it. Still, a single large facility with evaporative cooling can use over a million gallons on a hot day, and one 2026 estimate puts direct national water use at about 19 billion gallons a year, possibly rising to 60 to 110 billion by 2030.

People have noticed. In the first quarter of 2026 alone, Data Center Watch counted at least seventy-five projects that were blocked or delayed, worth about 130 billion dollars. Local opposition groups grew from 396 to 833 across forty-nine states, and lawmakers from both parties filed moratorium bills in fourteen legislatures.

If you look at these debates, you’ll see they are rarely about technical issues like latency or uptime. Instead, they are about what a place should be, what a community owes to the future, what can be taken from shared resources and with whose permission, and whether the promised benefits are real or just stories from outsiders. These are, in a strict sense, religious questions. They concern ultimate goals and which sacrifices are justified.

A society reveals its deepest values in what it chooses to build at great cost and declare essential.

Goryeo believed the canon was worth sixteen years of labor under siege. That conviction is carved into every block. Our own faith is etched into the substations rising around us.

The thing that could not answer

When people look for a religious story about artificial intelligence, they often turn to the golem, but most choose the wrong version.

The well-known story says Rabbi Judah Loew of Prague created a clay servant to protect his community, but then lost control of it. Loew died in 1609, and there is no evidence from his time or the next two centuries linking him to a golem. This connection first appears in German literature between 1834 and 1847. The story’s current form comes from Yudl Rosenberg’s Nifla’ot Maharal, published in Piotrków in 1909 and presented as a manuscript found in Metz. Gershom Scholem called it modern fiction. The tale we treat as ancient wisdom about machines surpassing their creators is actually a modern worry dressed up in old clothes.

The older sources ask a deeper question. In tractate Sanhedrin 65b, Rava creates a man and sends him to Rabbi Zeira. Zeira speaks to the creature, but it does not reply. Zeira then says, you were made by one of the companions; return to your dust.

The test was whether it could speak, not its strength, obedience, or threat. The creature was dissolved because it could not respond.

Think about that for a moment, in a time when machines can answer smoothly and at length. The tradition behind the golem story did not fear a talking creation. It judged personhood by the ability to respond, and it also included the strongest warning against trusting what people make. Psalm 115 criticizes idols for having mouths but not speaking. Ours do speak. The warning was never about whether the object works. It was about where people direct their worship.

Sacred vessels of the altar

Western Christianity offers the longest recorded example of people living thoughtfully with their tools, and it documented the experience.

The Rule of Saint Benedict, written in the sixth century, tells the cellarer to treat every tool and item in the monastery as a sacred vessel, leaving nothing uncared for. Cistercian monasteries used water-powered mills, forges, and fulling works, all organized around a schedule marked by hours of prayer. The Rule does not ask if a tool is efficient. It asks what the tool does to people’s time, to the shared table, and to the person using it.

Thomas Merton put that question to the twentieth century. In the summer of 1964, Ping Ferry sent him a manuscript of Jacques Ellul’s The Technological Society, published that year in John Wilkinson’s translation by Alfred A. Knopf and first issued in French in 1954. Merton filled three notebook pages, made three journal entries in a week, called it “great, full of firecrackers,” and then, days later, wrote that Ellul was too pessimistic though not unreasonably so. He reviewed it favorably in Commonweal that December. In his notes he left a line that has aged into a plumb bob: the attitude of the peace movement to technology will determine its spiritual base.

This is the Benedictine question for our age: not what the system achieves, but how it molds the lives of those who depend on it.

Nothing exists in isolation.

Buddhist analysis gives us the part most missing from our public debate, and it is not a warning but a description.

Dependent origination, or paṭiccasamuppāda, teaches that nothing exists on its own; everything depends on conditions. In this context, calling something ‘the model’ is just a convenient label. What really exists is a chain: a factory, a transmission line, a substation in a county with a tax break, an aquifer, a data annotator paid by the task, a collection of writing from unasked authors, a cooling tower, and a five-year depreciation schedule. Calling this chain a cloud is making a metaphysical claim, and it is not accurate.

Marshall McLuhan called this figure and ground. The chatbot is the figure, what everyone notices. The substation is the ground, the part that makes the figure possible but stays hidden. Religions are humanity’s oldest and most reliable way to make the ground visible. Practices like Sabbath, fasting, daily prayer, and giving alms all serve the same purpose: they break our focus on the figure so we can notice the ground. If a society cannot look away from the interface, it will miss the community it depends on.

The Mongols won anyway.

This argument deserves to be made clearly, because it is strong.

Religious institutions have often opposed new technologies. They resisted lightning rods, anesthesia in childbirth, dissection, and even printing in some places, always arguing about ultimate purposes—and they were wrong. Religious worldviews have justified resource extraction as often as they have limited it; reading Genesis as a call for dominion has cleared more forests than any spreadsheet. Bringing religious authority into infrastructure policy empowers groups with their own interests and no responsibility to those affected. The issues before a county board are truly technical, and a hydrologist is more helpful there than a preacher.

The strongest version of this objection is the story at the start of this essay. The canon was created to defend against the Mongols. The Mongols won.

All of that is true. The point isn’t that religions always understand technology. Goryeo lost the war. The real point is more specific and harder to ignore: religions are the largest surviving record of people deciding what is worth great sacrifice, who pays for it, what is owed to future generations, and what a built thing is really for. No other source has been tested over so many centuries and collapses. And it must be read across several traditions at once, so no single one has the final say. The goal is not authority, but a vocabulary big enough for the question.

What to ask before the concrete pours

1. Ask what the building is for before asking whether it works.
2. Look at the background, not just the main feature. Identify the substation, the aquifer, and the annotator. Find out by whom, and whether they agreed.
3. The old test of a made thing was whether it could answer. The new test is what we owe the people whose words taught it.
4. Judge any tool by what it does to the hours and the common life of the people who use it.
5. Watch where the worship goes.

Amazon told investors on 7 February 2025 that it had shortened the useful life of a subset of its servers from six years to five, effective 1 January 2025. That is the horizon on which the physical substance of this revolution is being planned: five years, then replacement.

The hall at Haeinsa has no wiring, no chillers, and no generators. Its engineering is a slope, two window sizes, and a floor of charcoal, lime, salt, and sand. It has held fifty-two million characters legible and mildew-free for more than five centuries, and in 1970 it beat a modern facility so decisively that the blocks came back up the mountain.

The people who built it were solving for a reader they would never meet.

That is the measure to aspire to.

The rest is procurement.

Sources

• UNESCO World Heritage Center, “Haeinsa Temple Janggyeong Panjeon, the Depositories for the Tripitaka Koreana Woodblocks” (inscribed 1995): https://whc.unesco.org/en/list/737/
• UNESCO Memory of the World Register, “Printing woodblocks of the Tripitaka Koreana and miscellaneous Buddhist scriptures”: https://media.unesco.org/sites/default/files/webform/mow001/korea_tripitaka.pdf
• “Haeinsa” and “Tripitaka Koreana,” Wikipedia (construction and rebuilding of Janggyeong Panjeon, ventilation and floor composition, the 1970 relocation attempt): https://en.wikipedia.org/wiki/Haeinsa
• Atlas Obscura, “Tripitaka Koreana” (1970 mildew and return of the blocks): https://www.atlasobscura.com/places/tripitaka-koreana
• “Goryeo under Mongol rule,” Wikipedia (capitulation of 1259, the killing of Choe Ui in 1258): https://en.wikipedia.org/wiki/Goryeo_under_Mongol_rule
• International Energy Agency, Energy and AI, executive summary: https://www.iea.org/reports/energy-and-ai/executive-summary
• Water Foundation, Water and Data Centers (2026), summarizing Han et al. (2026) and James et al. (2025): https://waterfdn.org/wp-content/uploads/2026/07/Water-Data-Centers-2026.pdf.
• Data Center Watch, Q1 2026 report: https://www.datacenterwatch.org/q1-2026
• Babylonian Talmud, Sanhedrin 65b, Steinsaltz edition: https://steinsaltz-center.org/portal/library/Talmud/Sanhedrin/chapter/65.b
• “Golem,” Wikipedia, on the nineteenth-century German origin of the Prague attribution and Yudl Rosenberg, Nifla’ot Maharal (Piotrków, 1909): https://en.wikipedia.org/wiki/Golem
• The Rule of Saint Benedict, chapter 31, on the cellarer: https://eriebenedictines.org/spirituality/rule-of-benedict.html
• International Jacques Ellul Society, “Ellul and Merton,” on Merton’s 1964 reading of Jacques Ellul, The Technological Society, translated by John Wilkinson (New York: Alfred A. Knopf, 1964): https://ellul.org/themes/ellul-and-merton/