The Lake They Couldn't See — Cyrus Radfar

Skip to content<br>Speed Read

Crater Lake and Wizard IslandSurvey photograph, 1874. Public domain.The deepest lake in the country, in one of its first photographs. A party hunting gold rode up to it in 1853, admired it for a few minutes, and rode away to keep digging.<br>Miners Working a Long TomCalifornia Gold Rush, c. 1852. Public domain.The diggers who chased the asset. They almost always lose; the man who sold them the shovels almost always wins. Every half-century the same play runs again.<br>Hydraulic MiningPhotograph, c. 1860s–70s. U.S. National Archives (296578).An iron water cannon dissolving a hillside. The monitors washed enough of the Sierra Nevada into the rivers to fill the Panama Canal excavation more than five times over, and the debris drowned the farm towns downstream.<br>The Great Eastern Laying the Atlantic CableHarper's Weekly, 1865. Public domain.A cable ship unspooling a line thinner than a hair across the floor of the ocean, steering blind on arithmetic. A century later the same impulse buried a continent of fiber, and barely a few percent was ever lit.<br>A Data-Center HallPhoto: BalticServers, CC BY-SA 3.0.The shovel of this rush. Windowless rooms of humming machines that run on electricity and water, fed by coal plants a utility had promised, in writing, to close.<br>A Klamath ManEdward S. Curtis, c. 1923. Public domain.The Klamath had known the lake the whole time, and held it sacred, and never needed to name it twice. The blindness was never real. It was selective, and a selective blindness has a beneficiary.<br>Crater Lake from the RimPhotograph. Wikimedia Commons, CC BY 3.0.Standing on the rim of the wave, trying to prepare for what comes after the dust settles. The lake was never lost. It was only worth nothing to men who were counting.

TL;DR First it was gold, then bandwidth, now machine intelligence. Each time, the people who chased the asset lost, the people who sold the shovels won, and nobody could see the things that carried no price. The AI data-center boom is the third run of the same play.

Crater Lake, Oregon. A party hunting gold rode up to it in 1853 and rode away. Photo: CGP Grey, CC BY 2.0.<br>On the morning of June 12, 1853, a prospector named John Wesley Hillman was riding a tired mule through the Cascade Mountains of southern Oregon, hunting a gold mine that may never have existed. His party had heard about the Lost Cabin: a vein so rich the men who found it had to abandon it and were never able to find their way back. So Hillman and a dozen others spent their days staring at the ground, reading the rock for color, half-mad with the fear that a rival party would get there first.

His mule stopped at the edge of a cliff. Hillman peered over the ledge and saw a lake unlike anything he had ever seen.

It would come to be known as Crater Lake. It is the deepest lake in the United States, nearly two thousand feet of clear water held in the throat of a collapsed volcano, and the water is a blue so saturated it looks poured in rather than reflected. The men admired it for a few minutes. They named it Deep Blue Lake. Then they got back on their mules and went looking for the gold.

The lake was promptly forgotten as the men went home to tell stories lacked any credibility. Two more parties stumbled onto it over the next sixteen years, each certain it was theirs to discover, before the name Crater Lake finally stuck in 1869. The Klamath had known it the whole time, and held it sacred, and never needed to name it twice. It was never lost. It was only worth nothing to men who were counting.

The shape of a rush

The diggers who chased the asset. They almost always lose; the man who sold them the shovels almost always wins. California Gold Rush, c. 1852.<br>The story isn’t about one surprising afternoon in the Oregon wilderness. It is a pattern that shapes every “gold rush”, and they come round again and again. Every half-century or so a new kind of gold appears, and we go take the world apart to get it.

Plenty of people will tell you that no longer holds, that progress now compounds too fast for fifty-year waves. The gaps are shrinking even in this essay: a hundred and fifty years between the gold and the fiber, twenty-six between the fiber and the machines. Anthropic’s Dario Amodei thinks powerful AI could fold a century of progress into a decade. He may be right that the inventing goes faster now. Inventing was never the part that took fifty years.

Deep Dive: Is the cycle actually speeding up? The accelerationists say yes, and have for a while. Ray Kurzweil’s law of accelerating returns holds that each advance shortens the road to the next, so the curve bends upward and the gaps between revolutions close; an AI that can improve AI would be the ultimate compression. Amodei’s “compressed century” is the strong form of the claim.<br>The stagnationists say the opposite. Robert Gordon and Tyler Cowen point out that measured productivity growth has been...