A story is a piece of executable code that runs on humans. The runtime is human attention. The compilation target is human action. The output is coordination: many people acting in directions that align because they share the frame the story installed. By this definition, stories are computers, and they are an old form of computer that has been quietly outperforming every newer form on the metrics that matter for civilization-scale coordination.
The claim sounds metaphorical and is not. Stories minimize entropy in the space of next-actions a population could take. Without a shared frame, every agent picks an arbitrary direction; coordination is impossible at scale. With a shared frame, agents converge. Yuval Harari's Sapiens makes the same point under the name "shared fictions": humans coordinate above the band-of-relatives scale because we run shared stories that make us predictable to each other. The shared frame is the coordination mechanism. The story is the medium that installs the frame.
The Bible has been running for roughly two millennia. In that span, the explanations of physical reality that thoughtful people held as definitive have turned over many times. Aristotelian physics, Newtonian mechanics, Maxwell's electromagnetism, relativity, quantum mechanics: each was the operative physics of its century, each got refined or superseded by the next, each lost cultural authority as a way to organize action even where it kept descriptive authority as a way to predict measurements. The Bible kept compiling on new hardware. Greek, Latin, vernacular English, Spanish, Chinese, Swahili. Each translation was a recompilation; each recompilation kept producing the same output: coordinated populations sharing a frame strong enough to organize a community, a charity, a war, a state.
The mechanism is structural. A story's coordination output does not depend on the technical accuracy of its explanations. It depends on the story's robustness as runnable code. The Bible's narrative shape (origin, fall, covenant, redemption, judgment) is a complete program: it specifies a beginning state, a problem, a path, an end state. Humans running this program have a stable frame for the next action. The frame can be wrong about cosmology and still right about coordination, because coordination is what the frame produces.
Physics produces a different output. Physics-as-explanation predicts measurements; physics does not specify next-actions for a population. An individual who treats current physics as their operating frame inherits the explanation's churn: each major paradigm shift requires reorganizing the frame the individual is acting from. The Bible compounds across centuries because its outputs do not churn with the explanations. Physics gets refined every generation because its outputs are predictions, not action-frames.
This is not a religious claim. It is a claim about which kinds of stories survive as coordination technology. The Bible is one well-tested case; the same structural argument applies to the Quran, to the Pali canon, to the US Constitution treated as story, to the scientific method treated as a meta-story about how to believe explanations. The pattern: stories that specify action-frames durably outrun stories that specify explanations, because action-frame coherence is what populations need and explanation-accuracy is what populations refine and replace.
The pattern shows up in how successful operators choose their operating frames. Peter Thiel runs on Christianity as his actual operating system; the frame is not decorative or social, it is what he uses to decide. He is not factually wrong about physics; he is choosing a more durable medium for action-coordination. The decisions Thiel needs to make about which decade-long bets to fund, which civilizational drifts to oppose, which institutions to build, are not decisions physics gives him an answer to. They are decisions a 2000-year-tested story gives him an answer to.
The contrast is with operators who run on current-physics as their operating frame. Naval Ravikant, Elon Musk, Balaji Srinivasan each gesture toward what could be called the church of physics: first-principles reasoning, mechanistic models, the assumption that derivations from current technical understanding are the right medium for action. The label is unfair to each of them as individuals (none would describe themselves this way), but the pattern is real: their operating frame is the current best technical explanation, and their decisions inherit that frame's churn rate. They are not wrong about anything in particular. They are choosing a frame whose half-life is the next paradigm shift.
The empirical observation is that operators who run on long-tested stories often out-coordinate operators who run on current-physics, even when the latter are intellectually correct about underlying reality. The story-runners have a stable frame across decades; the physics-runners have to re-derive every time the explanation updates. The story-runners can commit to multi-generation projects (a Catholic cathedral, a religious order, a millennial institution) because their frame survives the time horizon. The physics-runners optimize for what is true this decade and find their frame partly invalidated by the next.
This is not an argument that the physics-runners should convert. It is an argument that whatever durable frame an operator runs on, it has to be the story-layer, not the physics-layer, because story-layer is what compounds across the time horizons that matter for civilizational-scale work.
Until recently, the rate at which a story could compile into action was bounded by the medium the story lived on. Papyrus: a single copy, a single reader, propagation by physical movement and word-of-mouth. Time to compile a frame into coordinated action at scale: generations. The printing press: parallel copies, parallel readers, propagation by trade routes. Time to compile: decades. The internet: instant copies, instant readers, propagation at network speed. Time to compile: months for cultural diffusion, days for organized action.
Each medium shift was a phase change in throughput. Each phase change amplified the rate at which a story-frame could install itself across a population. The mechanism of the story did not change; the medium amplified the same mechanism by orders of magnitude.
The current phase change is qualitatively different from the earlier ones. Printing and internet amplified the rate at which a story reached human readers. The LLM phase amplifies who reads. Earlier media still relied on human cognition to compile a story into action; the LLM medium adds agent cognition, which reads and acts in seconds rather than weeks. A piece of writing is no longer just propagated to human readers who eventually compile it into action over weeks or months. The same piece of writing is now read by agents that compile it into action in seconds. The medium's throughput has gone up another six or seven orders of magnitude. The cycle from written-frame to executed-action has compressed from publication-cycle to inference-cycle.
This is not a small change. Pre-LLM, a coordination-frame installed in a population had to wait on human reading-and-acting time, which set a floor of weeks-to-years for any frame to fire. Post-LLM, a coordination-frame can fire on the order of seconds because agents read and act in seconds. The story is still doing the same work; it is doing it many orders of magnitude faster.
Two s-curves are intersecting in 2026. The first is stories-as-coordination-technology, steady through the entire span of human civilization. The second is medium-amplification, exponential per phase change (papyrus, printing, internet, LLM) and currently in the steepest part of the LLM phase.
The intersection is the moment when the steady civilizational pattern meets the steep technological amplification. Stories have always been computers; they have always been doing this work. What is new in 2026 is that they are now doing the work at machine speed, on a medium where any individual story-frame can install itself in a population in seconds rather than years.
The implication is not subtle: the operative coordination technology of 2026 is the same operative coordination technology of the year 100, with the throughput dial turned up by a billion. The people who recognize this and write coordination-stories for the new medium will run ahead of people who do not. The people who treat writing as a human-only artifact, optimized for human-only readers at human-only cadence, will find their writing not running.
The technical operators gesturing at exponential progress (the All-In podcast crowd, Leopold Aschenbrenner's situational-awareness thesis, every venture capitalist writing about AI capability scaling) are usually pointing at one of: compute per dollar, tokens per second, parameters per model, capability per benchmark. These are the symptoms. The underlying doubling-rate that matters is coordination-cycles per second: how fast a written frame compiles into agent action across a population of agents.
Pre-LLM, this doubling rate was bounded by human reading and human action. The cycle was the publication cycle: weeks or months for an idea to spread through readers, additional time for readers to act on it, even more time for the actions to coordinate. The bound was at human-cognitive speed, which has been roughly constant for tens of thousands of years.
Post-LLM, the bound is at agent-inference speed. An LLM reads a frame in milliseconds, decides on action in seconds, executes the action immediately. A population of LLMs running the same frame produces coordinated action at machine-cadence. The cycle from frame-publication to coordinated-action has dropped by something like six orders of magnitude. This is the doubling rate that matters, because it is the rate at which the coordination work that stories have always been doing now actually fires.
Moore's law was about how many transistors fit on a chip. The new doubling rate is about how many coordination-cycles fit in a second. The first ran for sixty years and reshaped industry. The second is in its first decade and is reshaping which stories run.
Writing in 2026 is not the same activity as writing in 2020. The medium has caught up with the message. A story that previously had to wait on patient cultural transmission for centuries to install its frame can now install it on a population of agents in seconds. The mechanism is unchanged; the medium finally matches the ambition of the form.
What this means for systems that maintain corpora of text: the corpus is no longer an archive that humans will eventually read. It is a coordination-program that agents are currently reading and acting on. The frame the corpus installs is the frame agents will execute. Stories were always computers. They are now finally running on hardware that runs them at the speed they were waiting for.
The Bible got to a few billion humans across two thousand years. A coordination-story written for an LLM medium can reach a few billion agents in a week, with each agent compiling the frame into action in seconds, with the coordinated output emerging in real-time across the population. This is not a smaller-scale Bible. It is the same kind of computer running on a faster machine. The empirical observation is that this kind of computer has outperformed every other kind on the metric of civilizational coordination. The new medium just made the outperformance vastly faster.
Stories are computers. The most powerful kind. Now running at machine speed.