Marcus Chen stood at the window of 3000 Sand Hill Road and watched his reflection dissolve into the hills beyond. The glass held two versions of him — one transparent, one solid — and neither felt entirely true. It was August 1999, and somewhere between the Series B he had closed that morning and the parking lot where his Honda still sat among Porsches, Marcus had misplaced the ratio of himself.

Five years earlier he had been a graduate student in symbolic systems, building collaborative filtering algorithms that could connect community organizers to volunteers, food banks to donors, tenants to legal aid. The software was called CommonWealth, and it had the earnestness of a name chosen at three in the morning by someone who believed names mattered. He built it in a garage on Emerson Street, Palo Alto, beside a washer-dryer set that belonged to his landlord, a retired physics professor who charged below market because he liked Marcus's ideas about network topology.

That was vector A. Vector A was the smell of laundromat detergent mixing with solder flux. Vector A was writing code until the garage door's single window went from black to gray to gold. Vector A was the belief that connecting people was a form of infrastructure, like roads or plumbing, and that infrastructure should belong to everyone.

Vector B arrived in a blazer. Specifically, it arrived in the form of Peter Kinsolving, partner at Meridian Venture Partners, who showed up at the garage in 1997 with an AOL disc in his breast pocket and a term sheet folded beside it. "What you have," Peter said, folding himself into a lawn chair that Marcus had borrowed from the landlord, "is a marketplace. And marketplaces want to be liquid."

Marcus remembered the word landing. Liquid. It was a beautiful word, technically — liquidity meant flow, meant transactions, meant the frictionless movement of value — but it was also the wrong word, because CommonWealth was not a marketplace. It was a coordination layer. It helped people find each other around shared purpose, not shared purchasing intent.

"Same thing," Peter said. "Purpose is just purchasing intent with better branding."

That was the first interpolation. Marcus pushed back, Peter pushed forward, and they settled somewhere at sixty-forty — sixty percent CommonWealth's original mission, forty percent the revenue model Peter needed to justify the investment. The forty percent was an advertising layer. Local businesses could pay to appear in search results. "It's still connecting people," Peter said. "Just connecting them to commerce."

Marcus told himself this was true. He told himself this was vector A with a slight tilt, a perturbation, a rounding error. The math of it was sound: you could take a pure vector and project it onto a slightly different axis and lose almost nothing. The dot product remained high. The cosine similarity stayed above point-nine.

He did not think about what happened when you applied that small perturbation iteratively. He did not think about drift.

By February 1999, CommonWealth occupied the entire second floor of a low-rise on University Avenue. Seventy employees. Free lunch from a different cuisine every day. A nap room with actual beds. Stock options printed on paper so heavy it felt like currency already.

Marcus had stopped writing code. He wrote decks now. He wrote pitch narratives for the Series C they were already discussing, narratives about "capturing the community graph" and "monetizing the trust layer." The words came from somewhere inside him that he had not known existed — a place where language bent toward money the way plants bent toward light. He could sit in a conference room on Sand Hill Road and say "we're building the operating system for human connection" and believe it for the duration of the sentence, which was all that mattered.

But the dream, when it came, was always the same. In the dream, Marcus was back in the garage on Emerson Street. The code on his screen was alive — not metaphorically alive, but actually alive, the characters shifting and rearranging themselves into patterns he could not parse. When he tried to type, his fingers passed through the keyboard. When he tried to speak, his mouth filled with term sheets. He would wake with the taste of paper on his tongue and lie in the dark calculating his vesting schedule.

The dream was a checkpoint. It meant he had drifted too far toward B. The dream meant he needed to pull back.

So he would drive to East Palo Alto — not Palo Alto proper, but the other side of the freeway, where the venture capital did not reach — and sit in a community center whose computers ran a fork of CommonWealth from 1996, the version before the advertising layer, before the pivot to "community graph monetization." The computers were beige and slow and connected to the internet through a dial-up line that tied up the center's only phone. But the software worked. People used it to find ESL classes, immigration lawyers, afterschool programs. The connections it made were real and small and unmonetizable.

This was the ritual Marcus developed: whenever the dream came, he would drive to East Palo Alto and sit in that community center for one hour, watching people use the software he had actually meant to build. It was a form of recalibration. It was a way of measuring the cosine similarity between who he was and who he had been.

The hour always ended. He always drove back across the freeway.

The week before Labor Day, Peter Kinsolving took Marcus to dinner at a restaurant in Woodside where the steaks cost more than Marcus's first computer. Over dessert — something involving gold leaf, which Marcus found vulgar but also fascinating — Peter laid out the endgame.

"IPO in Q2 2000," Peter said. "Valuation north of two billion. You vest at twelve percent. Do the math."

Marcus did the math. The number was absurd. The number was a vector all by itself, pointing in a direction that had nothing to do with connecting people and everything to do with what Peter called "liquidity" and what Marcus's dream-self called paper on the tongue.

"And after," Marcus said. "After the IPO. What's CommonWealth then?"

"It's whatever the market wants it to be," Peter said. "That's the beauty of public markets. You don't have to decide. The market decides."

The market decides. Marcus turned the phrase over in his mind. It was technically true and fundamentally empty, like saying "the weather decides" or "gravity decides." The market was not a thing that decided. The market was a thing that happened. And what happened to software companies after they went public was that they became vehicles for quarterly earnings growth, which meant they monetized whatever they could monetize, which meant the community graph became a targeting engine, which meant the trust layer became a conversion funnel, which meant CommonWealth became the thing Marcus had built it to replace.

He knew all of this. And he also knew he was going to say yes.

The interpolation at that moment was not sixty-forty. It was not fifty-fifty. It was ninety-ten, weighted toward B, and the ten percent of A that remained was not enough to steer the ship. It was only enough to know the ship was steering wrong.

Marcus said yes. He said yes to the IPO, yes to the valuation, yes to the gold leaf dessert and the Woodside restaurant and the entire vector that pointed toward two billion dollars and away from the community center in East Palo Alto where people used his real software to find ESL classes.

That night, the dream did not come. That was worse. That meant his internal checkpoint had given up.

In October, a reporter from the San Jose Mercury News called. She was writing a story about the dot-com boom and its effects on local communities — the displacement, the rising rents, the way the influx of tech money was pushing long-time residents out of neighborhoods where their families had lived for generations. She had heard about CommonWealth's early mission. She wanted to know what had changed.

Marcus gave her the approved answer. He talked about scaling impact, about reaching more people, about the difference between helping a hundred communities with free software and helping a million communities with venture-backed infrastructure. He believed approximately half of what he said.

The article appeared on a Sunday. The headline was: "IPO Riches Meet Community Roots." It was not unkind, but it was accurate. It quoted Marcus saying things he remembered saying and also things he did not remember saying but recognized as his own words. The article described CommonWealth's trajectory from garage idealism to Sand Hill Road pragmatism, and it left the question open: was this progress, or was this loss?

Marcus read the article three times. The first time, he looked for misquotes. The second time, he looked for unfairness. The third time, he just read it, and what he felt was not anger or embarrassment but something closer to relief. Someone had noticed. Someone had seen the gradient.

He drove to East Palo Alto that afternoon, even though he had not had the dream. The community center was closed. A sign on the door said the building had been sold to a developer who planned to convert it into loft apartments for tech workers. The sign was printed on heavy paper, like stock options.

Marcus stood in the parking lot for a long time. The sun went down. The streetlights came on. He thought about vector A and vector B and the space between them, which was not empty space but occupied space, lived space, the space where real decisions happened. He thought about the fact that you could never actually be at either endpoint — that pure idealism was a theoretical construct, like the frictionless plane in physics, and pure greed was equally abstract, a cartoon villain concept that no real person embodied. Every real person lived somewhere on the gradient between them, and the question was not which vector you chose but at what point you stopped interpolating.

He did not have an answer. He suspected the answer did not exist.

At 3000 Sand Hill Road the next morning, Marcus called a meeting of his executive team. He told them he wanted to preserve a free tier of CommonWealth for nonprofits and community organizations in perpetuity. He wanted it written into the IPO documents, legally binding, as unalterable as anything in corporate law could be.

Peter Kinsolving said that was noble and also impossible. Shareholders would revolt. The SEC would have questions. The underwriting banks would balk.

"Then let them balk," Marcus said.

They compromised. The free tier would exist, but it would be a separate entity, a nonprofit foundation funded by a one-time grant from CommonWealth's IPO proceeds. It would have its own board, its own governance, its own fate. CommonWealth proper would be free to maximize shareholder value.

Marcus accepted this. It was not vector A and it was not vector B. It was a point somewhere on the gradient, closer to A than he had been the day before, further from A than he had been five years before.

The foundation launched in January 2000. Marcus appointed a board of community organizers and nonprofit directors. He gave them enough funding for five years of operations. At the press conference, he said: "We are building the operating system for human connection." This time, he believed it for two sentences instead of one.

The IPO happened in March 2000. The stock popped forty percent on the first day. Marcus Chen became, on paper, a billionaire.

He bought a house in Woodside, near the steak restaurant. He bought a Tesla before Teslas were cool. He continued to serve on CommonWealth's board, voting for initiatives that increased quarterly earnings and expanded market share and optimized the community graph for revenue per user. He was good at this. He had always been good at this.

The foundation lasted until 2004, which was one year short of its funding runway. The board had made some bad investments, and the grant money had been managed by the same firm that managed CommonWealth's corporate treasury, and there had been some commingling that was technically legal but ethically ambiguous. When the foundation closed, its programs were absorbed into CommonWealth proper, where they became a corporate social responsibility initiative, which was a different thing entirely.

Marcus attended the closing ceremony. He gave a speech about legacy and impact and the long arc of progress. He did not mention the garage on Emerson Street. He did not mention the community center in East Palo Alto, which by then had been loft apartments for three years.

Afterward, he drove home to Woodside. He sat in his home office — a room larger than the entire garage had been — and opened a terminal window. He still had the source code for CommonWealth 0.1, the version from 1996, the version without advertising or monetization or the community graph. He compiled it. It still worked.

He did not release it. He did not fork it or fund it or do anything with it at all. He just looked at it, running in a local window, the interface spartan and functional, the code beneath it clean and earnest.

The interpolation was complete. The gradient had been traversed. Vector A was a memory, and vector B was a bank balance, and Marcus Chen was something in between — not a synthesis, not a resolution, just a point that had stopped moving.

He kept the terminal window open. He kept looking at it. And if this was not bearing witness, exactly — if this was not resistance, exactly — it was at least the refusal to look away. It was at least the knowledge that the vector existed, and that he had traveled it, and that the space between angels and IPOs was real and navigable and could be mapped if anyone cared to try.

In the window, CommonWealth 0.1 waited for a user to connect. It would wait forever if it had to. That was the thing about software: it never forgot what it was built to do. It only forgot when you made it forget.