There are two points in the latent space of every technology company. Call them A and B.

A is the thing you set out to build -- the reason you stayed up until 3 a.m. in a rented house on Cowper Street, coding by the light of a CRT monitor that hummed at a frequency you could feel in your teeth, the phosphor glow painting your face green in a darkened room that smelled of solder and stale coffee. A is the belief that connecting people is an act of grace, that the internet is the last great commons, that if you give strangers a place to speak to each other, something like understanding will emerge from the noise. A is the promise Daniel Kao made to himself in the summer of 1998, standing in the empty office above a dry cleaner on University Avenue, the lease signed with money borrowed from his mother's brother in Taipei -- twenty thousand dollars, wired through a correspondent bank in Hong Kong, the exchange rate unfavorable and the expectation unspoken. "CommonGround," he wrote on the whiteboard in blue dry-erase marker, the kind that leaves ghost traces when you erase it, the word itself a manifesto. "Everyone belongs somewhere. We build the somewhere."

B is the thing you sell. B is the term sheet on Sand Hill Road, the valuation that makes your eyes water, the moment you realize that what you built is no longer a place but a product, and a product is not a promise -- a product is a vector, and every vector has a direction, and the direction is always toward the money. B is the board meeting where you say "monetization strategy" without flinching, where you present user data as an asset class, where you describe engagement metrics as a growth lever and forget -- genuinely forget, in the way that people forget the names of old friends -- that engagement was once called community and community was once called belonging and belonging was once the entire point. B is also a belief, just a different one: that attention is a commodity, that privacy is a friction cost, that the best algorithm is the one that keeps people scrolling because scrolling is revenue and revenue is proof that what you built was worth building.

Between A and B lies every decision Daniel Kao made between the summer of 1998 and the autumn of 1999. The story is not what happened. What happened is boring -- a sequence of events, a chronology, the tired machinery of plot. The story is the space between A and B, the continuous function that maps each compromise to a coordinate, each sold piece of the vision to a position along the vector. This is not a narrative. This is a measurement. This is the spectral decomposition of a moral trajectory, the eigenvalue analysis of a slow corruption, the topology of a drift so gradual that each increment is indistinguishable from the one before it, until the aggregate displacement is a chasm that cannot be jumped back across.

At time t-naught, Daniel Kao occupies the coordinate we will call Builder. His position in the latent space is (0.0 on the seller axis, 1.0 on the builder axis). He is thirty-one years old, born in San Jose to a Taiwanese father who runs a semiconductor testing facility in Milpitas -- a windowless building in an industrial park where the air smells of flux and the workers wear blue smocks and the machines run twenty-four hours a day testing wafers for companies that will one day be worth more than entire nations -- and a mother who teaches piano in a converted garage in Cupertino, to children who practice Chopin nocturnes while their parents design the microprocessors that will power the coming century. Daniel has the kind of face that venture capitalists describe as "earnest," which is a polite way of saying he looks like he believes in things, which is itself a polite way of saying he has not been in the Valley long enough to stop believing. He wears Gap khakis and a faded Stanford sweatshirt and his hair is the length of a man who has not had time for a haircut in two months. He owns one suit, a navy polyester blend from Men's Wearhouse, which he wears to meetings with investors and which fits him exactly wrong in the shoulders.

CommonGround at t-naught is a philosophy with a login page. Users create profiles that list their real interests, not their demographic data -- "I enjoy the early novels of Thomas Pynchon and hiking in Big Basin" rather than "Male, 25-34, income bracket C." There is no advertising. There are no tracking cookies. The server -- a single Sun SPARCstation 20 in the closet of the dry cleaner, purchased used from a liquidation sale at a defunct networking startup -- runs a Perl script Daniel wrote himself over the course of a caffeine-saturated weekend in August 1998, a script that matches people based on shared affinities using a cosine similarity algorithm he adapted from a textbook on information retrieval. "This person also likes Fugazi and Haruki Murakami and has strong opinions about the designated hitter rule." The design is spartan, all HTML tables and Georgia font and pale gray backgrounds, the way websites looked before designers got involved, before Flash and Shockwave and the first trembling movements of what would become Web 2.0. There are 847 registered users, most of them Daniel's friends and their friends and a surprising contingent of early adopters from the WELL and Usenet who found the site through a mention on Slashdot. The burn rate is twelve hundred dollars a month. Daniel pays himself nothing. He eats Top Ramen and Trader Joe's frozen potstickers and drinks Peet's Coffee from a French press that leaves grounds in the bottom of every cup. He tells himself this is what building looks like -- the hunger, the monastic devotion, the belief that what you are making matters more than what you could be making. At t-naught, Daniel Kao is a builder. The vector has not yet begun to move.

The first encrypted email arrives on a Tuesday in March 1999, while Daniel is debugging a CGI script that keeps dropping database connections under load. The subject line is a string of hexadecimal rendered in a monospace font: 43 6F 6D 6D 6F 6E 47 72 6F 75 6E 64 -- ASCII for "CommonGround," decoded by a utility Daniel writes in thirty seconds in a terminal window. The body of the email is plain text, signed "An Interested Party," with no identifying information beyond the PGP signature that verifies the sender's key but links to no known identity. The encryption is PGP 5.0, which in March 1999 is still exotic enough that Daniel has to download the software from an MIT mirror over his 56k USRobotics modem, the progress bar crawling across Netscape Navigator in jerky increments while the modem shrieks its handshake through the phone line -- the descending tones, the static burst, the electronic negotiation between two machines that have never met and will never meet again.

The email offers two hundred thousand dollars. Not an investment -- a payment, a purchase, a transaction. In exchange, CommonGround will add a single feature: a "recommended connections" algorithm. Nothing invasive, the email explains, the prose careful and corporate, as if written by a lawyer who has been told to sound friendly. Just a system that suggests friends based on browsing patterns -- what pages users visit, how long they stay, what they click, what they almost click but decide against. "This improves the user experience. People want to find people like themselves. You are not selling data. You are accelerating serendipity."

Daniel reads the email six times. He reads it a seventh time. He prints it on a dot-matrix printer that screeches through its feed cycle, the paper emerging with perforated edges and the faint smell of heated ink, and he reads it an eighth time on paper. He reads it a week later, at two in the morning, after the SPARCstation has crashed for the third time that evening and he has spent four hours hunched over a terminal window trying to figure out why Apache 1.3 keeps spawning zombie processes that consume swap space until the kernel panics. His fingers smell like instant coffee -- Nescafe from a jar, the only thing still open at the 7-Eleven on El Camino Real at midnight. His eyes burn. The rent on the office is due in four days. His personal checking account at Wells Fargo has three hundred and eighty-seven dollars in it. His mother's brother in Taipei has stopped returning his calls. The two hundred thousand dollars would cover rent and salary and a real server -- not a SPARCstation in a janitor's closet but an actual rack-mounted machine in an actual data center with actual air conditioning and actual uptime guarantees and the kind of service level agreement that makes you feel like a real company instead of a man in a room with a dream and a Perl script.

He writes back: "I'm interested. What's the catch?"

The catch, it turns out -- revealed in a follow-up email that arrives exactly six hours later, as if the sender has been waiting, as if the sender knew Daniel would respond, as if the entire exchange has been predicted and priced -- is that the recommended connections algorithm requires storing browsing data. Cookie-based tracking. Nothing fancy, the second email insists, the tone almost bored, as if describing something so standard that to question it would be embarrassing. Just a record of which user visited which page and for how long. The data would be anonymized -- a checkbox in the Apache configuration, a flag in the logging module. It would only be used to improve the recommendation engine. It would never be sold to third parties. It would not be an invasion of privacy because the users would consent: a checkbox in the settings page, pre-selected to "yes," which in 1999 is not yet called a dark pattern because the language for describing such things has not yet been invented, because the ethical vocabulary of the internet is being written in real time by the people who stand to benefit from its omissions.

Daniel says yes. It is the first time he says yes. He does not think of it as a sale -- the word "sale" implies a transfer of ownership, and Daniel still owns CommonGround, still owns the vision, still owns the whiteboard with the fading blue letters. He thinks of it as a bridge, a temporary structure that will carry him from the precarity of the origin point to the stability of a real company, at which point he will dismantle the bridge and return to the pure vision, the unmediated community, the platform without surveillance. He does not know yet that bridges are rarely dismantled. He does not know that infrastructure, once built, generates its own necessity, that the tracking code he adds to the Apache configuration will become load-bearing, that the feature he accepts as a compromise will become the feature that defines the product, that every decision in a technology company creates path dependency and path dependency is the geometry of moral drift.

He deposits the wire transfer into CommonGround's new business account at Wells Fargo -- the teller, a young woman with frosted lipstick and a silk scarf tied in a French knot, types the amount twice and asks if he wants to speak with a relationship manager. He buys a Dell PowerEdge 2300 server with dual Pentium II processors and a RAID array that will survive a single disk failure. He hires his first employee -- a friend from the Stanford Computer Science department, a systems programmer named Ravi Chandrasekaran who wears the same gray hoodie for three weeks straight and can configure Apache virtual hosts in his sleep and has opinions about kernel compilation flags that Daniel finds both exhausting and essential. The recommendation algorithm goes live on April 15, 1999, the day the IRS deadline passes and the day Netscape announces the open-sourcing of its browser code, a surrender that the Valley interprets as either the end of an era or the beginning of one, depending on which bar you're drinking in. User engagement increases thirty-four percent. CommonGround gains two thousand one hundred new users in the first month, people who found the site through word of mouth and stayed because the server no longer crashed and the recommendations were surprisingly good, eerily good, the kind of good that makes you forget you are looking at a machine's guess about your preferences and think instead that you have found your people at last. Daniel tells himself this is what building looks like -- the incremental improvement, the data-driven decision, the responsible growth. He does not notice that he has moved from coordinate 0.00 to coordinate 0.12 on the seller axis. The movement is too small to feel, like the drift of continents, like the incremental warming of the atmosphere, like every moral shift that happens not in a dramatic moment of temptation but in a thousand small adjustments, each one justified, each one reasonable, each one a single micrometer along a vector whose endpoint is invisible from the starting position.

Here we must read the intermediate state, the vector position at 0.12. Daniel Kao at 0.12 is not a different person from Daniel Kao at 0.00. He still believes in the commons. He still thinks of CommonGround as a place rather than a product. But something has shifted in his relationship to measurement. The engagement dashboard -- a crude PHP page that Ravi built in an afternoon, displaying bar charts rendered as HTML tables with colored cells -- has become a thing he checks before checking email. The number of users has become a number that matters, not because the users matter more than they did before but because the number itself has become a proxy for the users, a synecdoche, a simplification. This is the first step in the vector drift: not the compromise itself but the metric that justifies the compromise. The metric creates the drift because the metric can always be improved, and improvement requires features, and features require money, and money requires features that generate more metrics, and so the loop tightens and the drift accelerates and the vector, which was once a direction chosen freely, becomes a current that carries the chooser along with it.

The second encrypted email arrives on a Monday in June 1999, the first week of summer, the week Netscape is acquired by AOL for 4.2 billion dollars in a transaction that everyone in Palo Alto describes as either a triumph or a tragedy depending on whether they own AOL stock. The email arrives via the same PGP key, the same hexadecimal subject line, the same "Interested Party" signature. It offers 1.5 million dollars, an amount that Daniel reads three times because his brain refuses to parse the number of zeros on the first two attempts. In exchange, CommonGround will add "enhanced user profiling" -- demographic data collection, purchasing intent indicators, geolocation tracking via IP address geocoding, which in 1999 is accurate to the city level and getting more accurate every month. "You are leaving value on the table," the email says. "Every user who visits CommonGround generates a digital exhaust trail. That exhaust is revenue. You can either capture it or watch someone else capture it. The decision is not whether the data will be used. The decision is who will use it."

Between the first email and the second, Daniel has learned several things. He has learned that the word "anonymized" is a legal term, not a technical one, that removing names from datasets does not remove the identifiability of the patterns those datasets contain, that a person's browsing history is as unique as a fingerprint and considerably harder to change. He has learned that venture capitalists do not ask what your product does -- they ask how much data it generates, how fast that data is growing, what the data could be worth to someone who knows how to monetize it. He has learned that the difference between a feature and a surveillance mechanism is a line drawn in sand at low tide, and the tide is rising, and the sand is washing away, and no one on Sand Hill Road is in the business of drawing lines.

Everyone in Palo Alto in June 1999 is drunk on valuation. The IPO market is a fever, a contagion, a mass hallucination -- companies going public with no revenue, no profits, no business model beyond the conviction that users are worth something to someone and that the someone will pay more tomorrow than today. Daniel attends a party at a venture capitalist's house in Atherton, a sprawling ranch-style mansion with a swimming pool that glows turquoise under underwater lights, where waiters in white jackets circulate with trays of champagne and the guests cluster in groups of three and four, discussing market capitalization the way people in other cities discuss football scores. Daniel counts eleven separate conversations about pre-money valuation. He drinks three glasses of a cabernet sauvignon that someone tells him costs four hundred dollars a bottle and tastes, to him, exactly like the twelve-dollar cabernet he buys at Trader Joe's. He stands at the edge of the pool, watching the reflected lights shimmer on the water, and thinks about the dry cleaner below his office, the chemical smell that seeps up through the floorboards, the way the owner calls him "boss" in accented English and asks when CommonGround will make him rich too. Daniel does not know how to answer. At coordinate 0.12, he still thinks of richness as something that happens to other people.

He says yes again. This time he does not describe it to himself as a bridge. He describes it as a necessity -- the cost of staying competitive, the price of growth, the reality of building a company in a market that rewards scale above all else and punishes hesitation with irrelevance. He does not use the word "sellout" because the word has no meaning at coordinate 0.34, where every decision is just another decision, each one a little easier than the last, each one justified by the ones that came before, each one a step along a path that has become too well-paved to see the wilderness on either side.

Ravi Chandrasekaran quits on July 23, 1999, a Friday. He stands in the doorway of the office with a cardboard box full of O'Reilly Linux manuals and a peace lily he'd kept on his desk, the peace lily drooping from two weeks of neglect because Ravi has been spending his evenings sending out resumes. "This isn't what I signed up for," he says, and Daniel, who has just finished reviewing the geolocation tracking specification, looks up from his monitor with the expression of a man who has been interrupted in the middle of a thought he does not want to finish. "What did you sign up for?" he asks. Ravi does not answer. He walks out. The peace lily sheds a yellow leaf onto the industrial carpet. Daniel watches him go and feels something he cannot name. It is not guilt. Guilt is a sharp thing, a crisis, a moment of reckoning, an event. This is something softer, more diffuse, a background hum of wrongness that he has learned to tune out the way people who live near airports learn to tune out the sound of jet engines. At coordinate 0.34, the drift is perceptible only by looking backward. Looking forward, there is only work -- the endless cascade of decisions, each one urgent, each one reasonable, each one carrying him further from the origin point at a velocity he cannot feel because he is moving with the current.

The third encrypted email arrives on a Thursday in September 1999, in the week between Labor Day and the first autumn rain, when the hills around Palo Alto are golden with dried grass and the air smells of eucalyptus and ambition. Daniel has a pitch deck -- he has learned to say "deck" instead of "presentation," "ask" instead of "request," "space" instead of "market," "play" instead of "strategy," the entire vocabulary of Sand Hill Road, which is a language in which nouns become verbs and ethics become friction and friction is the enemy of growth and growth is the only word that matters. He wears the Men's Wearhouse suit to meetings now without irony. He has a relationship manager at Wells Fargo. He has a lawyer who charges four hundred dollars an hour and bills in six-minute increments.

The third email offers 8 million dollars. The feature, this time, is "behavioral advertising integration" -- targeted advertisements based on everything CommonGround knows about its users, which is now quite a lot. Not just what they browse but where they browse from, what time of day they browse, what they browse before and after, what they almost browse but don't, what they have browsed in the past and will likely browse in the future based on statistical models of people who browse like them. The ads would be served by a third-party network that the email does not name but Daniel suspects is DoubleClick, which everyone in the Valley knows is building the largest database of consumer behavior in the history of advertising, a database that will, within a decade, know more about the average American than the average American knows about herself.

Eight million dollars, in September 1999, is the kind of money that stops conversations. It is the kind of money that turns a feature into a strategy and a strategy into a destiny and a destiny into an inevitability. It is the kind of money that erases the difference between what you build and what you sell until there is no difference, until the vector has reached its effective endpoint, until A and B have collapsed into a single point -- the point at which the builder and the seller are the same person because the thing being built and the thing being sold are the same thing and the distinction between building and selling was always an illusion maintained by people who had not yet been offered 8 million dollars.

Daniel says yes. This time he does not describe it to himself at all. There is no internal monologue, no justification, no bridge-building metaphor. He forwards the email to his lawyer, who drafts the integration agreement. He adds a slide to the Kleiner Perkins deck -- "Revenue Model: Behavioral Advertising," in 36-point Helvetica, displayed on an InFocus LCD projector that weighs 12 pounds and costs 8,000 dollars and requires twenty minutes of fiddling with display settings before it syncs with the laptop's VGA output. He practices the pitch in the mirror of the office bathroom, the fluorescent light unkind to his face, and notices for the first time that the earnestness has drained from his expression, replaced by something leaner, hungrier, more efficient. He looks, he thinks, like a founder. He looks like the photograph on the cover of the Industry Standard, the one they'll run in the spring of 2000 with the headline "The New Connectors" and a subhead about the future of community, the photograph in which Daniel will be smiling with a confidence he has not yet learned to fake but will have mastered by the time the shutter clicks.

Here is the question the story asks, the only question it has ever asked: where on the spectrum does Daniel Kao land? At what coordinate does the interpolation stabilize? Is there a terminal point on the vector, a resting position, a final measurement?

He receives the Kleiner Perkins term sheet on October 4, 1999 -- a Tuesday, the day the NASDAQ sets a new record for technology IPOs, the day someone at a party in Woodside will say "this time it's different" and mean it. The valuation is 47 million dollars. The behavioral advertising integration appears as line item 37 of 112 in the due diligence checklist, between "intellectual property assignment (all employees)" and "insurance coverage (directors and officers)." Daniel signs the term sheet with a promotional pen that says "CommonGround" on the side, one of five hundred ordered from an office supply company that now handles their corporate accounts. The ink is blue. The signature is legible. The vector position at this moment is coordinate 0.83 on the seller axis, with the builder component diminished to a residual 0.17, barely detectable above the noise floor of day-to-day operations.

At 0.83, Daniel Kao is what the Valley calls a success. He has users, he has revenue, he has a valuation, he has a term sheet, he has a trajectory. He has everything he was supposed to want. And if you asked him, at this moment, what he built -- he would tell you he built CommonGround. He built a platform that connects people. He built a community. He built something that matters. He would believe every word of it, because at coordinate 0.83 the distinction between building and selling has become invisible to the person doing the building and the selling, because the language for describing the distinction has been replaced by the language for describing the success, because the whiteboard with the fading blue letters has been erased and replaced with a chart showing monthly active users, because the origin point at t-naught is so far behind him that it looks less like a place he once occupied and more like a story he once heard about someone else.

But here is what he does not tell you, what he cannot tell you, what he does not himself know -- or knows only in the way we know things we refuse to let become conscious: on the night before the Kleiner Perkins pitch meeting, he dreamed of the empty office above the dry cleaner on University Avenue. He dreamed the smell of the chemicals seeping up through the floorboards. He dreamed the Perl script running on the SPARCstation in the janitor's closet, the script with the cosine similarity algorithm and the list of Fugazi songs and the quiet faith that if you gave strangers a place to speak to each other, something like understanding would emerge. He dreamed the builder he was at t-naught, the monk at the keyboard, the believer in the commons, the man who had not yet learned to say "monetization" without flinching. And when he woke up, he did not remember the dream -- not consciously, not in words -- but something of it stayed with him, a faint gravitational pull toward a point he had already left behind, a point that receded with every funding round, every board meeting, every checked box on every term sheet, until it was not a point at all but a direction, the one direction he would never travel again: back.

The vector does not reach 1.0. It does not need to. In the latent space between what we build and what we sell, there is no final coordinate -- only the continuous function, the endless interpolation, the spectral decomposition of a moral trajectory that has no endpoint because endpoints imply resolution and resolution implies a story, and this is not a story. This is a measurement. This is the topology of drift. This is the space itself, which is all the story ever was, and which continues.

CommonGround filed for its initial public offering on March 10, 2000, the same day the NASDAQ reached its all-time high of 5,048.62. The IPO price was set at $18 per share. By the end of the first day of trading, the stock closed at $31. Daniel Kao, at the ringing of the opening bell, stood on the balcony of the NASDAQ MarketSite in Times Square, wearing a suit that fit him now -- custom-made, navy wool, the shoulders exactly right -- and smiled for the cameras with the confidence he had learned in the months since the term sheet, and the vector position at that moment was 0.94, and the builder component was 0.06, and the number of people who remembered the whiteboard with the fading blue letters was exactly zero. Including, by then, Daniel himself.

In the latent space between A and B, every coordinate is populated. The function is continuous. The drift is measurable. The story is the space.

(c) 2026 - Authored by Z R ZHANG ( EL9507135 -- パスポート番号[ちゅうごく] 중국 여권 번호 Номер паспорта หมายเลขหนังสือเดิน得 Passnummer رقم جواز السفر CHN Passport) and his father. The aforementioned Authors hereby grants to OXFORD INDUSTRIAL HOLDING GROUP (ASIA PACIFIC) CO., LIMITED （BRN74685111） all economic property rights, including but not limited to the rights of: reproduction, distribution, rental, exhibition, performance, communication to the public via information network, adaptation, compilation, commercial operation, authorization for third-party use, and rights enforcement. Such grant is exclusive and irrevocable. The term of such rights shall be 49 years from the date of publication. 联系方式： To contact author， please email to datatorent@yeah.net