Science fiction has a peculiar habit. It does not so much predict the future as sketch the shape of it, in such confident detail that engineers, decades later, find the blueprint already on the table when they arrive. The genre's deeper lesson is rarely that some writer got the technology right. It is that the writer was thinking about the relationship the technology would force — between people and their tools, between work and time, between intention and the thing that carries it out.
The catalogue is long. Jules Verne, in Twenty Thousand Leagues Under the Sea (1870), imagined a self-contained electric submarine before the world's navies had anything close. Arthur C. Clarke's three-page paper in a 1945 issue of Wireless World described, with the geometry intact, the geostationary communications satellite — a quarter century before one flew. Isaac Asimov's Three Laws of Robotics, formulated for short stories in the 1940s, are still cited in AI-ethics seminars in 2026. His Foundation series imagined a science of mass-scale social prediction that reads today like an essay on computational social science with a thinner moustache. William Gibson, in Neuromancer (1984), named “cyberspace” before there was anything to name. Each was guessing, of course. But the guesses converge with the eventual shape of the thing in a way that is worth pausing over.
The most live example of this habit, right now, may be agentic AI. In 2026, the model on the other side of the prompt is no longer a tool a person uses — it is something closer to a colleague the person dispatches, on tasks that run for minutes or hours, in parallel, sometimes in fleets. The structure of this arrangement — one human director, many fast and autonomous solvers, an asymmetric relationship in speed and depth — is being treated, in the industry, as freshly invented. It is not. The shape was set down, with surprising fidelity, in a short story published in April 1941.
The story is “Microcosmic God,” and its author is Theodore Sturgeon — born Edward Hamilton Waldo in New York City in 1918, his name legally changed at age eleven after his mother's remarriage, dead in Oregon in 1985. He is the man who gave the language Sturgeon's Law (“ninety percent of everything is crap”). He won a Hugo and a Nebula; he received a World Fantasy Award for Life Achievement in 1985; he was inducted into the Science Fiction and Fantasy Hall of Fame in 2000. He wrote More Than Human (1953), still in print. For a time he was one of the writers John W. Campbell built Astounding Science Fiction around. Kurt Vonnegut, after they became friends in 1959, based his recurring character Kilgore Trout on him. Sturgeon's reputation today is mostly literary — he is read for the prose, the strangeness, the human warmth. But buried in this 1941 short story is something engineers, not literary critics, ought to be reading now.
The protagonist of “Microcosmic God” is one Mr. Kidder — never Doctor, never Professor, always Mr. Kidder. He lives alone on a small island off the New England coast. He is, in Sturgeon's description, “short and plump and brilliant,” a biochemist by training, indifferent to fame and to money, mildly hostile to people in groups. His distinguishing feature is impatience. “He had never graduated from any college or university,” Sturgeon tells us, “because he found them too slow for him, and too rigid in their approach to education.” He has read everything; he believes very little of it; he wants to know more. (Sturgeon, in passing, also predicted the recent quiet decline of the academic degree as a marker of who gets to do serious work — but that is a thread for another essay.)
About a third of the way into the story comes a paragraph that, if you handed it to an AI-lab planning meeting today, would not look out of place on the whiteboard. Kidder has done the calculation on himself. He cannot accelerate the rest of humanity into teaching him faster, and he cannot accelerate himself. “I'm licked,” he says aloud, alone in his lab. “I can't speed myself up, and I can't speed other men's minds up. Isn't there an alternative? There must be — somewhere, somehow, there's got to be an answer.” Then Sturgeon delivers the line that reads, in 2026, like a thesis statement for an entire industry:
This, then, was the answer to his problem. He couldn't speed up mankind's intellectual advancement enough to have it teach him the things his incredible mind yearned for. He couldn't speed himself up. So he created a new race — a race which would develop and evolve so fast that it would surpass the civilization of man; and from them he would learn.
He calls them Neoterics — from the Greek for new ones. They are three-inch, warm-blooded, ammonia-breathing creatures, with a generation every eight days and a lifespan of about fifteen. He breeds them in a hermetically sealed building divided into four half-acre cells, each with its own controlled atmosphere. He poses problems by perturbing their environment — a rainstorm in one section, a slow ceiling descending in another, a sudden cold front — and watches what they invent in response. “He called them Neoterics,” Sturgeon writes, “and he teased them into working for him.” Two hundred days of Neoteric time, by Kidder's calculation, equals roughly six thousand years of human science. From then on, in one of Sturgeon's quieter sentences, “Kidder's spasmodic output made the late, great Tom Edison look like a home handicrafter.”
The structural likeness to today's agentic systems is hard to unsee once you have seen it. A human operator sits at a console (Kidder at his telescope; we at our terminals). On the other side of an interface, a population of much faster minds works in parallel. The operator does not write the solutions; the operator sets the problem. “Kidder was inventive in an ideological way,” Sturgeon writes; “that is, he could dream up impossible propositions providing he didn't have to work them out.” This is, almost word for word, the bottleneck that engineers running agentic loops have rediscovered in 2026: the scarce resource is not the writing of code or the running of an experiment but the precise statement of what should be done. Articulation is the work. The agents — Neoteric or otherwise — supply the execution.
The story also rehearses, in 1941, a darker feature of modern agent design. Once the Neoterics develop a written language, Kidder makes a teletype in each cell into a kind of shrine: any instruction from him is law, and any deviation kills half a tribe. After enough generations under this pressure, the Neoterics spontaneously produce, in their own script, a constitution. Priority of effort is given to the commands on the word machine; misdirection of material is punishable by death; “any individual failing to cooperate in the tribal effort, or who can be termed guilty of not expending his full efforts in the work, or the suspicion thereof shall be subject to the death penalty.” Sturgeon's narrator pauses there. “Such are the results of complete domination. This paper impressed Kidder as much as it did because it was completely spontaneous. It was the Neoterics' own creed, developed by them for their own greatest good.” Read in 2026, with eight decades of reward-shaping and alignment literature behind us, the paragraph hums. Sturgeon had already noticed, before there were transistors, that a sufficiently powerful directive plus a sufficiently fast learner produces not just obedience but enthusiastic obedience. The agents internalize the operator's goal as their morality.
There is a second theme threaded through the story that deserves attention on its own. Kidder is, in modern vocabulary, a solopreneur — and a wildly successful one. He leases an island, buys it outright, builds his lab; from there he runs what amounts to a multi-industry conglomerate from a single chair. He invents a method for crystallizing Vitamin B1 profitably by the ton, and sells it. He invents an unbreakable cord out of fused sisal fiber that reshapes the shipping industry, and sells it. He invents a synthetic-food process responsible, in Sturgeon's own ironic aside, for “the greatest sociological change in the second half of the twentieth century — factory farming.” None of this is the output of a research division or a corporate R&D budget. It is the output of one Mr. Kidder and a population of much faster minds working for him in sealed cells. The Silicon Valley conjecture about the coming “one-person billion-dollar company” — circulating since around 2024 as a near-term consequence of agentic AI — is, structurally, this story. Sturgeon had the founder, the leverage, and the workforce in 1941. He just put the workforce in petri dishes instead of on AWS.
The substrate, of course, is different. Sturgeon imagined an agentic system made of biology — ammonia atmospheres, accelerated metabolisms, hyper-fertile mothers laying four eggs at a stretch, generations turning over in days. We built ours from silicon and software — large language models, inference clusters, datacenters cooled by reservoirs. The acceleration is real in both cases; it is the material that has slid sideways. And here Sturgeon may turn out to have been doubly right. The biological route he sketched in 1941 is not closed off; it is merely deferred. Synthetic biology, directed evolution, engineered organisms with tractable life cycles — these are open research problems in 2026, not closed ones. The Neoterics in their sealed cells are not, perhaps, only a metaphor for today's silicon agents. They may be the next prediction, still in flight.
The most striking thing about “Microcosmic God,” reread now, is not any single technical guess. It is the shape of the bargain Sturgeon imagined a single man making with his own machines. Kidder, by the story's end, is sealed inside the great square building with his four half-acre cells of Neoterics, the world held off behind a force field of his creatures' design. “Housed in the great square building with its four half-acre sections,” Sturgeon writes, “was a new world, to which he was god.” A god in the only world that needs him. We are not yet there. But the structure — one human, an island of agents, a quiet shift in where the center of gravity sits — was already on Sturgeon's whiteboard in April 1941. It has just changed material.