Frame 37: 1997, January. The garage at 214 Everett Avenue, Palo Alto. Two-thirty in the morning. The space heater has been running for six hours and the room smells of hot metal and cold ambition. Mira Kinsley is twenty-five years old, sitting cross-legged on a concrete floor, surrounded by whiteboards covered in equations that look like the scribbling of a mad god. The equations describe a search algorithm — not a better way to find web pages, but a fundamentally different way to understand what a question means. Most search engines at the time match keywords. Mira's algorithm maps semantic space, measuring the distance between concepts the way astronomers measure the distance between stars. She has been working on it for eighteen months, living on ramen and the dwindling remainder of her Stanford fellowship, sleeping on a futon in the corner of the garage because the apartment she was supposed to be renting turned out to be a sublet scam and she never bothered to find a new one. The whiteboards are covered in math that no one else has ever written. The code, when she runs it at 3:14 AM on a laptop with a cracked screen, returns results that make her cry. Not because they are perfect — they are far from perfect — but because they demonstrate something she has believed since she was sixteen and first read about Shannon's information theory in a public library in Des Moines: that meaning can be measured, that understanding can be quantified, that the distance between two ideas is not a metaphor but a number. She calls the algorithm Lumina. She calls it that because the first search she runs is for the etymology of the word "illuminate," and the algorithm tells her it comes from the Latin luminare, to light up, and she thinks: yes, that is what I want to do, I want to light up the space between what people ask and what they actually mean. She saves the code to a floppy disk and writes "Lumina v0.1" on the label in Sharpie. She puts the disk in a ziplock bag and tucks it into the pocket of her hoodie, the same hoodie she has been wearing for four days, and she falls asleep on the futon with the disk pressed against her heart like a promise she has made to herself.

Frame 122: 1999, March. The conference room at Kleiner Perkins Caufield and Byers, Sand Hill Road. The table is made of something that might be mahogany or might be walnut; Mira cannot tell and does not care. She is wearing a blazer that she bought at Goodwill and a pair of shoes that pinch her toes, and she is trying to explain Lumina to six men in blue shirts who keep asking about market capitalization and exit strategies and whether she has considered pivoting to enterprise software. The algorithm has evolved. It is no longer a research project. It is a company, incorporated in Delaware, with twelve employees and an office above a coffee shop on University Avenue and a valuation that makes Mira's head spin when she thinks about it. But the core idea remains the same: semantic search, understanding questions rather than matching keywords, mapping the space between what people mean and what they type. The men in blue shirts do not understand the math. Mira knows this within the first three minutes of the pitch. What they understand is the growth curve — the user numbers that have doubled every month since August, the engagement metrics that make the existing search engines look like card catalogs in a burning library, the demo she shows them of Lumina answering a query about "the thing that killed the dinosaurs" by returning a page about the Chicxulub impact crater even though the user never typed the word "asteroid." That is the magic. That is what the men in blue shirts see, and that is what they want to own. The term sheet arrives three days later. Forty million dollars, Series A, at a valuation of two hundred million. Mira signs it in her office above the coffee shop, using a pen that someone has given her as a gift, a Montblanc with gold fittings that feels wrong in her hand. She signs and she thinks: this is how it starts, the compromise, the dilution, the slow drift away from the origin point. She thinks: I will not let it happen. She thinks: I will stay true to the vision. She thinks this while the ink is still wet on the paper, and she believes it, and that is the tragedy.

Frame 218: 1999, November. The launch party at the Fairmont Hotel in San Francisco. Champagne in crystal flutes. Oysters on silver platters. Three hundred guests, half of them investors, half of them employees, all of them wearing the expression of people who have bet on the right horse and know it. Mira stands at the edge of the ballroom, watching the celebration as if from behind a pane of glass. She is twenty-seven years old, and she is worth, on paper, four hundred and eighty million dollars. The algorithm is powering search for seventeen million users. The word "Lumina" has become a verb — "I'll Lumina it" — and the press calls her a visionary, a genius, the woman who taught machines to understand meaning. She should be happy. She is not happy. The champagne tastes like someone else's celebration. The oysters slide down her throat like pieces of the future she is already losing. Because the algorithm is no longer hers. Not really. The Series A gave the investors two seats on the board, and the board has started asking about advertising, about user data monetization, about an acquisition offer from a media conglomerate that wants to turn Lumina into a portal with weather widgets and stock tickers and banner ads for mortgage refinancing. Mira has fought every battle. She has won some of them. But she can feel the vector shifting, the trajectory bending, the interpolation point sliding away from the origin coordinates she set in a cold garage in 1997. She leaves the party early. She walks back to her hotel room — she does not have an apartment anymore, she has not had a permanent address in eight months — and she opens her laptop and looks at the original code, the Lumina v0.1 that she still keeps on a floppy disk in a ziplock bag in her suitcase. The code is beautiful. It is clean. It is pure. It is also irrelevant, because the algorithm that is running on seventeen million devices is a different thing entirely, a thing that has been optimized for speed and scale and the particular needs of the advertising infrastructure that the board has already decided to build. She closes the laptop. She does not sleep. She sits in the dark and watches the lights of San Francisco flicker through the fog and thinks about the distance between what she meant and what she has become.

Frame 304: 2000, July. The crash has come. Not slowly, not gently, but like a trapdoor opening beneath the feet of everyone who believed that the laws of gravity had been suspended for the duration of the dot-com boom. The NASDAQ has lost forty percent of its value since March. Lumina's valuation has been cut to a third of its peak, and the board is in emergency mode, and the CEO they brought in six months ago — a professional manager from Oracle with a tan and a handshake that feels like a contract negotiation — has proposed a plan: pivot to enterprise search, license the algorithm to corporations for internal knowledge management, abandon the consumer product entirely. Mira opposes the plan. She is outvoted six to one. The CEO sends her an email — a single sentence, no greeting, no signature — saying: "The market has spoken." Mira reads the email seven times, each time hoping the words will rearrange themselves into a different message, and then she closes her laptop and walks out of the office and does not come back for three days. She spends those three days in the garage at 214 Everett Avenue, which is still there, which has been preserved like a shrine to the origin point, the whiteboards still covered in equations, the space heater still in the corner, the futon still flattened from the weight of a twenty-five-year-old woman who had once believed that meaning could be measured and understanding could be quantified and the distance between ideas was not a metaphor but a number. She sits on the concrete floor and she cries. Not for the money — the money was never real, the money was a shared hallucination that everyone had agreed to believe in for as long as it was convenient. She cries for the algorithm. She cries for the seventeen million users who trusted Lumina to tell them the truth about the world. She cries for the girl from Des Moines who had walked into a public library and read a book about information theory and understood, for the first time, that the universe could be explained in numbers.

Frame 401: 2001, February. The acquisition has closed. A division of a Fortune 500 company has purchased Lumina's technology for seventy-three million dollars, a fraction of its peak valuation but enough to return capital to the investors who had insisted on the enterprise pivot. Mira is no longer employed by the company she founded. She has a non-compete agreement that prevents her from working in search for four years, and a bank account with twelve million dollars after taxes, and an apartment in the Mission District that she has furnished with nothing but a mattress on the floor and a bookshelf that holds the floppy disk containing Lumina v0.1. She does not know what to do with herself. She has never not been working on Lumina. The algorithm has been the organizing principle of her existence for four years, the coordinate system by which she has oriented every decision, and without it she is floating in an undefined space with no origin to measure from. She starts attending conferences again, not as a speaker but as a spectator, sitting in the back rows of auditoriums listening to other people talk about the future of search, the future of artificial intelligence, the future of meaning itself. The presentations are polished and professional and completely hollow, and Mira sits in the darkness and thinks about the distance between what people say and what they actually mean, and she realizes that this was what she had been trying to measure all along, and she had succeeded, and now she was living inside the gap she had discovered.

Final Frame: 2001, December. Mira walks into the Stanford computer science building for the first time in five years. She is not a student anymore. She is not a founder anymore. She is something else — a vector that has been pushed so far from its origin that it no longer recognizes its own trajectory. She finds an empty classroom on the third floor, a room with a chalkboard and a single window overlooking the quad, and she sits at a desk in the back row and opens a notebook that she has carried with her since 1997. The notebook contains the original equations for Lumina. It contains notes from the garage, from the pitch meetings, from the board meetings where she fought and lost. It contains the record of every interpolation point along the vector from origin to endpoint, from the girl who wanted to measure meaning to the woman who had sold her measurement tool to a corporation that would use it to serve ads for diet pills and mortgage refinancing. She does not cry. She does not rage. She takes out a pen — a cheap ballpoint, the kind she had used in the garage, not the Montblanc with the gold fittings — and she writes one final entry in the notebook. She writes: "The distance between Vector A and Vector B is not a failure of the algorithm. It is what the algorithm was designed to measure. The error was not in the calculation. The error was in believing that the distance could be crossed without losing what was on the other side." She closes the notebook. She leaves it on the desk. She walks out of the classroom and down the stairs and out into the December rain, which is cold and clean and smells like the eucalyptus trees that line the campus pathways. She does not know where she is going. She does not know what she will do next. But she knows the distance now. She has measured it, precisely, and the measurement itself is a kind of truth — the only truth that survived the interpolation. The algorithm endures. It powers search engines she will never use, recommendation systems she will never see, advertising platforms that sell things she will never buy. It has been refined, optimized, stripped of its original purpose and rebuilt for profit. But somewhere in the code, buried under layers of patches and optimizations and the fingerprints of a hundred engineers who never saw the garage or the whiteboards or the futon, there is a kernel of the original idea: that meaning can be measured, that understanding can be quantified, that the space between what people ask and what they actually mean is not a void but a vector. Mira Kinsley walks through the rain and does not look back. The algorithm continues to run. The distance between the origin point and the final frame can be calculated to seven decimal places. The meaning of that distance cannot be calculated at all.