The Sand Hill Road offices of Kleiner Perkins smelled like money and new carpet. David Chen sat in the walnut-and-leather waiting room, watching the receptionist answer calls on a phone that probably cost more than his first car. Through the glass wall he could see the partners' conference room, where men who had funded Netscape and Amazon and Excite@Home were now deciding whether NeuraLogic was worth eight million dollars. The air conditioning hummed at exactly seventy-one degrees, the optimal temperature for decision-making. David had read that somewhere, probably in Wired magazine, probably while pulling an all-nighter at the Stanford computer lab where this whole thing started.

He was twenty-nine years old. His mother in Taipei still called every Sunday to ask if he had eaten. His fiancee Priya was doing her residency at Stanford Medical, working hundred-hour weeks, and they saw each other mostly in the twenty minutes between when she collapsed into bed and when he left for the office at four in the morning. The engagement ring sat in its velvet box in his desk drawer because they had not found time to plan a wedding. David carried a prototype of NeuraLogic's neural recommendation engine on a Zip disk in his backpack, right next to a half-eaten PowerBar and the business cards he had printed at Kinko's three months ago because he could not afford proper letterpress.

The vector had two poles. David had been staring at them for six months, ever since the first term sheet arrived.

Pole One: What NeuraLogic could be. The technology was genuinely elegant, a collaborative filtering algorithm that learned user preferences across media types. David had built the first version to help his mother find Taiwanese-language audiobooks on the early web. He had refined it to recommend academic papers to fellow graduate students. In the right hands, NeuraLogic could connect people to content that educated them, challenged them, expanded their worlds. It could be the librarian of the internet, the curator that helped users discover what they needed rather than what was merely addictive. When David talked about this version of the company, his voice changed. Even the VCs noticed. They leaned forward slightly in their Aeron chairs, because somewhere beneath their cynicism they remembered being young and believing that technology could make the world better.

Pole Two: What NeuraLogic was being pushed to become. The Sand Hill Road playbook was simple and brutal. Acquire users by any means necessary. Optimize for engagement, which was a polite word for addiction. Sell user data to advertisers. Build recommendation loops that kept people clicking, scrolling, refreshing, never leaving. The venture capitalists did not say "addiction platform" out loud. They said things like "maximizing session duration" and "reducing churn" and "creating sticky experiences." But David was not stupid. He had read B.F. Skinner. He understood variable reward schedules. He knew that the most valuable algorithm in the world was the one that figured out exactly which emotional trigger would keep a specific human being glued to their screen for one more minute, and then another, and then another, until the evening was gone and their children were asleep and they had consumed nothing but algorithmically optimized content designed to make them feel just satisfied enough to never leave but just empty enough to always want more.

Between these two poles ran the vector, and David Chen was being asked to pick a point along it.

The conference room door opened. John Doerr's assistant, a brisk woman in Ann Taylor, gestured him in. David stood, straightened his tie—a gift from Priya, navy silk, she had said it made him look like someone who deserved funding—and walked into the room where the future of NeuraLogic would be determined.

The partners wanted Pole Two. They did not say it in those words, but David could read the subtext in every question. How would you monetize the recommendation data? What partnerships do you have with e-commerce platforms? How quickly can you get to ten million monthly active users? Can you beat Amazon's recommendation engine to market? The word "market" appeared seventeen times in the forty-five-minute meeting. The word "user" appeared forty-three times. The word "human" did not appear at all.

David answered their questions. He was good at this, had always been good at performing competence, the immigrant son's particular talent. He described NeuraLogic's technical architecture in precise detail, walked them through the user acquisition funnel, projected revenue curves that looked like hockey sticks because all dot-com revenue curves looked like hockey sticks in 1999. The partners nodded. John Doerr actually smiled once, a rare event that David's friends in the Valley had told him was worth approximately two million dollars of additional valuation.

But all through the meeting, David was thinking about Rahul.

Rahul Mehta, his co-founder, his best friend since the second week of Stanford's computer science orientation. Rahul who had written the first lines of NeuraLogic's code on a whiteboard in a Palo Alto coffee shop while David sketched neural network diagrams on napkins. Rahul who could debug a memory leak at three in the morning and still make you laugh about it. Rahul who had designed the database schema that could scale to millions of users, who had built the server infrastructure from spare parts and eBay auctions, who had mortgaged his parents' condo in Cupertino to keep the company alive during the three months when they had zero revenue and all the ramen they could eat.

Rahul who, three weeks ago, had been diagnosed with early-onset Parkinson's disease.

He was thirty-one years old. The neurologist at UCSF had been gentle but unambiguous. The disease would progress. There were treatments, expensive treatments, treatments that could slow the degeneration, preserve motor function for years, maybe a decade or more. But those treatments cost money. The kind of money that did not come from a startup founder's salary. The kind of money that came from an acquisition, an IPO, or both.

Rahul had not asked David to sell out. He had not asked for anything. He had told David the diagnosis, sat silently for maybe thirty seconds, and then opened his laptop and started writing a unit test for the recommendation engine's Bayesian inference module. That was Rahul. That was always Rahul, the man who solved problems with code because code was the one thing in the universe that behaved predictably.

But David had done the arithmetic. He had done it at two in the morning in the empty office on University Avenue, the spreadsheet glowing pale blue on his PowerBook screen. The numbers were not ambiguous. If they took the Kleiner term sheet, the one that demanded an engagement-maximization strategy, the one that would turn NeuraLogic into an addiction platform, they could IPO within eighteen months. Rahul's shares would be worth eight figures. Enough for any treatment, any experimental protocol, any clinical trial anywhere in the world. Enough to buy time, maybe enough time for a cure.

If they refused, if they held out for a funding round that allowed them to build the ethical version of NeuraLogic, they might not survive. Ethics were not a revenue model. The dot-com bubble was already showing its first cracks, the NASDAQ climbing toward a peak that everyone on Sand Hill Road pretended not to see. A company that refused to maximize engagement was a company that would be outcompeted, outspent, and eventually out of business.

The vector ran between Rahul's life and David's integrity, and every point on that vector was a different kind of betrayal.

If David took the money and built the addiction platform, he would be betraying every principle he had ever claimed to believe in. He would be building the thing he had always said he would never build. He would go home to Priya at night and know that his workday had been spent making the world slightly worse, slightly more hollow, slightly more optimized for emptiness.

If David refused the money and tried to build the ethical version, he would be betraying Rahul. The math was the math. Without the Kleiner money and the addiction platform, Rahul would not get his treatments. The disease would progress. David would watch his best friend deteriorate, would watch the tremor spread from Rahul's hands to his arms to his voice, would watch the man who had built NeuraLogic's foundation lose the ability to type, to speak clearly, to live independently.

There was no middle ground. David had searched for it, had spent sleepless nights hunting for a third option the way he used to hunt for buffer overflow exploits in his early hacking days. But the third option did not exist. Either Rahul's health or David's integrity. Pick one.

The Kleiner Perkins offices emptied. John Doerr shook David's hand and said something encouraging about "synergies" and "runway" and "category-defining opportunities." David walked out onto Sand Hill Road, the California sun blinding after the air-conditioned darkness, and stood on the sidewalk watching venture capitalists climb into their Porsches and Mercedes. Somewhere in Palo Alto, in a rented house on Waverley Street, Rahul was probably still writing code, his fingers steady for now, his mind still sharp, the clock ticking.

David had not told Priya about the diagnosis. He had not told her about the Kleiner term sheet, not in detail. She was too tired when she came home, her scrubs still smelling of hospital antiseptic, her eyes hollow from watching people die. She was a doctor. She saved lives. David built software. What was software, measured against a human life?

He walked to his car, a dented Honda Civic that he had bought from a graduating senior three years ago and never bothered to replace because cars were just things, and sat in the driver's seat without starting the engine. The leather was hot from the sun, the steering wheel almost too hot to touch. He thought about his mother in Taipei, calling every Sunday, asking if he had eaten. He thought about the Zip disk in his backpack, carrying the prototype that could become either a librarian or a dealer. He thought about Rahul's hands, steady for now, and about the tremor that would come.

The vector interpolation was not a choice between good and evil. It was a choice between two different kinds of harm, and David Chen was the one who had to pick the magnitude and the direction.

He could tell himself that addiction platforms were inevitable. Someone would build them eventually. If not NeuraLogic, then someone else, someone with fewer scruples, someone who would not even pause to consider the ethics. Better to have a conscience at the wheel of the machine than to let it be driven by pure profit motive. This was the argument for Pole Two, and it was not entirely wrong.

He could tell himself that Rahul would never want David to compromise his principles. Rahul was a good man, a principled engineer, the kind of person who commented his code and documented his APIs and never cut corners. Rahul would rather face his disease with dignity than force his best friend into moral bankruptcy. This was the argument for Pole One, and it was not entirely right either, because Rahul was also a man with a mother who called him every Sunday, a man who wanted to live, a man who had poured his savings and his parents' savings into a startup that was supposed to be their shared dream.

David started the engine. The radio came on, KZSU Stanford's student station, playing something electronic and melancholy. He pulled out of the parking lot and drove east on Sand Hill Road, past the venture capital offices that held the fate of every startup in the Valley, past the Stanford campus where he had first learned that code could change the world, past the hospital where Priya was probably elbow-deep in someone's open chest cavity, past the rented house where Rahul was writing code with hands that would not be steady forever.

The vector had no comfortable points. Every position along it was a concession, a compromise, a calculation in which human life and moral principles were measured in the same cold units. David had always believed that technology existed to serve humanity. He had believed it the way he believed in gravity, as a fundamental truth that required no defense. But now he was learning that service could mean many things. It could mean giving people what they needed. It could mean giving people what they wanted. And in 1999, at the peak of the dot-com boom, what people wanted was increasingly indistinguishable from what would destroy them.

He pulled into the driveway of the Waverley Street house. The lights were on in the living room, the blue glow of monitors visible through the windows. Rahul was inside, building the future, one line of code at a time.

David sat in the car for a long moment, the engine idling, the radio playing something he did not recognize. He thought about the Zip disk in his backpack. He thought about the term sheet from Kleiner Perkins. He thought about the vector and the two poles and the arithmetic of souls.

Then he opened the door and went inside to tell Rahul what he had decided.

The conversation took forty minutes. David laid out the options, the vector, the two poles. He explained the Kleiner term sheet, the engagement-maximization strategy, the addiction platform that Pole Two required. He explained the alternative, the ethical version, the slower growth curve, the very real possibility of failure. He did not mention Rahul's diagnosis, did not say "I am making this decision to save your life," because that would have been cruel in a way David was not yet prepared to be.

Rahul listened without interruption. When David finished, Rahul sat in silence for what felt like a very long time. The refrigerator hummed. A moth beat against the window screen. Somewhere in the house, one of their servers clicked and whirred, processing data, learning preferences, building the recommendation engine that would change the world one way or another.

Then Rahul said, quietly, "Take the money."

David started to protest, started to argue that they could find another way, but Rahul shook his head. The tremor was barely visible, just a slight unsteadiness in the gesture, but David saw it.

"You think I don't know why you're really considering this?" Rahul said. "You think I haven't done the same arithmetic?" He smiled, but it was not a happy smile. "You're trying to save my life by selling your soul. That's the deal. That's the equation."

"It's not that simple."

"It is exactly that simple. You are my best friend. You are the most principled engineer I have ever met. And you are standing in front of me, asking permission to betray everything you believe in, so that I can afford treatments." Rahul's voice did not break, but it came close. "I do not want to die, David. I am thirty-one years old and I do not want to die. But I also do not want to be the reason you become the kind of person who builds addiction platforms."

Neither of them spoke for a while. The moth kept beating against the screen. The servers kept processing data.

Finally David said, "I don't know what the right answer is. I don't know if there is a right answer."

Rahul looked at him for a long moment. "There isn't. That's the point. There are only vectors and positions and whatever point you pick, someone gets hurt. The question is whether you can live with the point you picked."

David thought about Priya, about the engagement ring in his desk drawer, about her hands saving lives while his hands built algorithms. He thought about his mother in Taipei, calling every Sunday, asking if he had eaten. He thought about the future, about the year 2000 and the year 2001 and all the years after that, about what kind of man he would be when he looked back on this moment.

"Okay," David said. "Okay."

He did not say which okay. He did not specify which pole he was choosing, which point on the vector he was picking. Maybe he did not know yet. Maybe the decision was still being computed, deep in the neural networks of his own mind, the weights adjusting and readjusting as new data arrived. Maybe the vector was not a line between two points but a space, a latent space, where every position was a different version of the man he could become.

Rahul opened his laptop. David opened his. The servers hummed and clicked and processed. Outside, the California night was warm and clear, the lights of Palo Alto spreading toward the hills, the dot-com boom still climbing toward its peak, the future still unwritten, the arithmetic still incomplete.

David began to code. There was a term sheet to respond to, a platform to build, a company to grow. There was a friendship to preserve and a fiancee to marry and a mother to call on Sunday. There was a vector to travel, a distance to cover, and at the end of it—wherever the end was—a man who would have to live with whatever point he had chosen along the way.