[Night without a date, late autumn 1999. The apartment above the garage. 427 Emerson Street, Palo Alto.]

The rain in Palo Alto doesn't make anything new. It just washes the pollen off the Volvos and the Beemers and makes the palms on University Avenue look like they've been polished by a detailer with nothing better to do. I sat on the floor of my apartment—the apartment above the same garage where I'd started a company four years and three funding rounds ago—and watched water streak down the single window. My hands were shaking again. Not the good kind of shaking, not the caffeine-jitter coding-all-night shaking. This was something deeper, tectonic, like a server rack about to fail and nobody's watching the monitoring dashboard.

The amber vial sat on my nightstand. The one Foster Keane had handed me after the Series B closed, along with a Montblanc pen and a speech about how "our best portfolio founders optimize every variable." Vitalis Protocol. Experimental. For founder use only. Four pills a day. The prescription wasn't signed by a doctor—it was signed by Foster. His signature, his pen, his pills, his founder. Every part of me belonged to the term sheet, including the parts I couldn't see in the mirror anymore.

I opened the vial. Thirty-two pills left—I'd been counting, the way you count the days until something ends or begins, or both at the same time. They spilled into my palm, amber capsules that glowed like fossilized honey in the dim light. Thirty-two tiny teeth, each one a bite of the person I used to be.

[Dawn, February 1995. The roof of the Gates Computer Science Building. Stanford University.]

Amir and I sat on the cold Spanish tiles, legs dangling over the edge, watching the sun come up over the Dish. We'd been up all night in the lab, not working on our dissertations—working on The Thing. We didn't have a name for it yet. We had coffee from the Treehouse in paper cups gone cold and a whiteboard marker that was running out of ink and an idea that felt like it might actually change something.

"What if we built something that couldn't be owned by anyone?" Amir said. He was chewing on a stale bagel from the CoHo. "Something truly decentralized. People own their own data. Identity is self-sovereign. No one company controls the protocol. No server farm in the middle where all the power sits."

I was twenty-four years old. I believed in things. I believed in open protocols and the Cathedral and the Bazaar and the idea that code could be a kind of constitution. We'd both read Lessig and Stallman and Raymond. We'd sat through Tim Berners-Lee's talk at MemAud and walked out convinced that the web didn't have to become what Netscape was turning it into—a walled garden with a gift shop at the exit.

"We'll call it Mesh," I said. "Because it connects everyone without a center."

The sun hit the red tile roofs of the Quad and everything looked golden. Everything looked possible. Amir smiled—he didn't smile often, which made it mean something when he did—and wrote three words on the whiteboard with the dying marker: NO CENTRAL AUTHORITY.

[Afternoon, June 1998. Conference room at 3000 Sand Hill Road. Benchmark Capital.]

The room smelled like new carpet and old money. I was wearing a blue blazer Amir had picked out at the Men's Wearhouse on El Camino Real—forty-eight dollars, and it looked like it. Foster Keane sat across the glass conference table in a charcoal Zegna suit that cost more than my parents' Honda Accord, tapping a gold Cross pen against his teeth. The tap-tap-tap was metronomic, a countdown to a question I didn't have an answer for.

"Mesh gives people control of their digital identity," I said. I'd practiced this pitch in the mirror of the garage bathroom, the one with the rust stain under the faucet. "Users own their data. It's a protocol, not a platform. Decentralized by design. Like email—no one owns email."

Foster stopped tapping. The silence was worse than the tapping. "How do you make money?"

The question hit me like a null pointer exception. I had an answer about API licensing and enterprise support tiers—B2B recurring revenue, scalable, high margin—but it came out sounding like something I'd read in a Harvard Business Review my dentist kept in his waiting room. Foster saw the hesitation and smiled the smile of a man who'd found the crack in the hull.

"We like the vision," he said. "The vision is good. The vision gets you on the cover of Wired. We'll give you eight million dollars. Series A. But at some point—and you know this, David, you're smart enough to know this—a protocol has to become a product."

I nodded. I'd been in the Valley long enough to know that nodding was the first step toward compromise, the way a zipper starts at the bottom and doesn't stop until everything's open. Amir, in the chair next to me, did not nod.

[Early morning, March 1995. The garage. 427 Emerson Street.]

Concrete floor cold enough to feel through sneakers. The smell of motor oil from the previous tenant's Datsun 240Z, still lingering after three weeks of scrubbing with Simple Green. Two folding tables from Office Depot. A Cisco router Amir had smuggled out of the Stanford networking lab in his backpack. A whiteboard covered in equations and network topology diagrams that made sense only to the two of us—and even then, only when we were both in the room.

We wrote the first lines of Mesh at 3:47 AM on a Tuesday. I remember the exact time because Amir typed it into the CVS commit message: 3:47 AM. Genesis commit. Let there be light.

The code was lean. No telemetry hooks. No user tracking. No "engagement metrics." No database collecting email addresses for a newsletter we'd never send. Just a peer-to-peer identity protocol written in C, with a cryptographic handshake that verified nodes without trusting any central authority. When a new machine joined the network, it didn't phone home to any server. It just joined. Everyone was a peer. Everyone was equal.

"One day," I said, "people will look back at this and say we built the infrastructure of freedom. Like the people who built TCP/IP."

Amir didn't say anything. He was already writing the next function, fingers moving across the keyboard like they were reading Braille, translating an architecture only he could see. That was the thing about Amir—he built with his hands, not his mouth. I was the one who talked. I was the one who pitched. Maybe that was the problem from the beginning. The talker always ends up selling what the builder built.

[Evening, December 1997. The Garden Court Hotel on Cowper Street. Mesh hits one hundred thousand registered users.]

Champagne flutes on silver trays carried by waiters in black vests. Venture capitalists in Ralph Lauren polo shirts and Allbirds before Allbirds existed. Engineers who'd never worn a tie wearing ties they'd borrowed from their roommates. A banner in Myriad Pro that read 100K USERS — THE MESH IS GROWING.

We'd pivoted. The word "pivot" was just entering the Silicon Valley lexicon in 1997—a euphemism for "we discovered our original idea couldn't generate revenue so we became something else entirely." The decentralized protocol had become a platform. "Better user experience," we called it. "Simplified onboarding." "Reducing friction." But what we really meant was: centralized servers in a colo in Santa Clara, user databases in MySQL, login credentials stored in plaintext for the first six months until an intern from Caltech noticed and nearly quit. The mesh had a center now. The center was us.

I gave a speech I didn't remember writing. The PR firm had drafted it—Summit Strategies, Susan Wojcicki's old shop before Google poached her, seventy-five hundred a month plus expenses—on letterhead with an embossed logo. I read it from a teleprompter that someone had set up behind the last row of chairs, the words scrolling past like a stock ticker. "Mesh is about connecting people. We're building community. We're building a better internet."

Someone put a champagne flute in my hand. I didn't like champagne. I drank it anyway, the bubbles burning the back of my throat. Across the room, Amir stood by the window, staring out at the palm trees on Cowper, not talking to anyone. He'd already told me he was leaving. "This isn't what we built," he'd said. "You know that, David. You know exactly what we built and what this is."

I knew. I'd known since the term sheet.

[Afternoon, September 1999. Corner office above University Avenue. The Series C term sheet.]

Forty-five million dollars. Kleiner Perkins, Sequoia Capital, and Keane Ventures. Pre-money valuation: two hundred and eighty million dollars. My equity after three rounds of dilution: four percent, fully diluted, with a four-year vesting schedule and a one-year cliff that would reset if we hadn't IPO'd by Q2 2000. Joseph from Wilson Sonsini had explained it to me with flowcharts and a laser pointer. By the time he was done, I understood that I owned less of my company than Foster Keane's administrative assistant owned of his.

I was reading the fine print—the stuff below the boldface, the stuff nobody reads—when I found it. Section 14.3, subsection (c): "Founder shall adhere to the Vitalis Protocol as directed by the Lead Investor for the duration of Founder's employment with the Company. Non-compliance shall constitute Cause for termination with forfeiture of all unvested equity and clawback of previously vested shares at the Board's discretion."

My body was in the term sheet. My brain chemistry was a line item between "Intellectual Property Assignment" and "Non-Compete Obligations." I'd signed away the thing that made me—the twenty-four-year-old on a cold roof at dawn who believed that code could be a constitution—and I hadn't even noticed when the pen touched the paper. The signature looked like mine. The hand that made it didn't.

I pulled the amber vial from my desk drawer. "Cognitive optimization," Foster had called it. "All our top founders are on it. Keeps you sharp for the twenty-hour days. It's what separates the unicorns from the also-rans."

The side effects: insomnia, dermatological lesions, memory fragmentation, parasthesia, cardiac arrhythmia. Long-term effects: unknown.

Unknown. The most expensive word in the English language. In Hollywood it was worth more than gold. In the Valley it was worth the entire market cap of the NASDAQ.

[Morning, October 1999. The Mesh boardroom. 165 University Avenue, third floor.]

"User engagement is up forty percent month-over-month," Caitlin was saying. She was the VP of Growth, hired from McKinsey at three times my salary. Everything she said sounded like a PowerPoint slide. "Our recommendation algorithm—the one David's team built, great work David—is increasing session length by an average of twelve minutes per user per day. Users who see targeted advertising convert at three point two times the rate of the control group. The click-through data is feeding back into the personalization engine, which means we're getting smarter every day. We're not just serving ads. We're building a preference graph of every user on the platform."

Foster leaned back in the Herman Miller Aeron chair he'd insisted on ordering for the boardroom. The same gold pen. The same tap-tap-tap. "The users are the product, David. You understand that now, don't you? That's what I was trying to tell you back at the Series A. You don't make money from users. You make money from knowing things about users that they don't know you know."

I nodded. I'd been nodding a lot lately—it had become my default gesture, a autonomic response like breathing or blinking. My hands were under the table because they were shaking, the deep tectonic shaking from the Vitalis, and the lesions on my arms itched under my long-sleeved Zegna shirt. I'd told the company doctor it was stress. He'd written me a prescription for more Vitalis.

"You're doing great," Foster said. "The IPO roadshow starts in January. Goldman's leading the book. You'll be a billionaire before you're thirty, David. Just keep optimizing."

I was twenty-eight. The age I'd been when we closed the Series A. The age I'd be forever, preserved in the amber of Foster's pills, an eternal startup founder, a permanent tech genius, a product that never aged because it was never alive to begin with.

[Night in a month that could be any month. The Mesh data center. Page Mill Road, near the old HP campus.]

Rows of Sun Microsystems E4500 servers, eighteen racks in a climate-controlled room that hummed like a cathedral organ, each machine singing its own frequency and together making a note just below hearing. 4.7 million user profiles—names, locations, IP addresses, browsing histories, purchase histories, private messages that had never been private, all of it living inside these black metal boxes with their rows of blinking green LEDs.

I walked between the racks. My reflection moved with me, a ghost in the dark glass of each server door—the same face that had been on the cover of Red Herring last spring, the same face from the Wired photoshoot where they'd made me stand in front of a green screen and pretend to type code I'd written years ago. The face of the future, they'd called it. The face of Web 2.0. The face of a product manufactured by a venture capital firm and sold to users who thought they were buying community.

I'd built this. Every line of code. Every database schema. Every API endpoint. The recommendation engine, the preference graph, the targeted advertising platform—my hands had typed it, my brain had dreamed it, my sleepless nights had fed it like fuel. And now it was a surveillance machine so elegant that the users didn't even know they were being watched. They thought they were part of a community. They thought the mesh connected them to each other.

The mesh connected them to us. To me. I was the center of the mesh. I was the spider in a web no one could see because the web was designed to be invisible.

My skin itched. The lesions had spread from my arms to my chest and down my spine. I'd stopped looking in the mirror without a shirt. The Vitalis was keeping me "optimized"—awake, focused, productive, the perfect founder for the perfect IPO—but something was eating me from the inside out. The same thing that had eaten the garage idealism, the clean code without telemetry, the belief that what we were building mattered more than what it could sell for. The machine consumes everything. Including its maker.

An email had arrived that morning. From Amir. The first in over a year. "I'm building something new," it said. "Open source. No VCs. No term sheets. No Vitalis. The way it was supposed to be. You could come help, David. There's still time."

I had read it seven times and hadn't replied. What would I have said? That I was four percent of a company I'd founded? That my body belonged to Keane Ventures? That the only thing I'd built worth building was the thing I'd spent four years and forty-five million dollars dismantling?

[Midnight. The apartment above the garage. November 1999. Or maybe it was always November. Maybe it had been November since the day I signed the term sheet.]

The rain had stopped. Outside the window, the palm trees on Emerson Street stood motionless in the dark, the way California palms do after a storm passes, like they're holding their breath, like they're waiting for something they know is coming.

I sat on the floor with my back against the wall—the same wall that had held the whiteboard covered in Amir's equations, the same floor with the oil stain from the Datsun 240Z, the same building where a twenty-four-year-old had once typed Let there be light into a CVS commit message and believed it with his whole body. The amber vial was empty now. Thirty-two capsules, not four. The honest dose. The only number I'd chosen for myself since the Series A closed.

I thought about the term sheet. About Section 14.3(c). About the founders Foster had mentioned—"all our top founders are on it"—and the ones who weren't top anymore. The founder of NetGravity who'd checked into a facility in Sedona. The CTO of ClickPath whose LinkedIn profile had frozen at September 1997 and never thawed. The Stanford roommate of a Stanford roommate who'd built a company, raised a round, started the Vitalis protocol, and disappeared so quietly that even Google couldn't find him anymore.

Amir was right. There was still time. Just not the kind of time the term sheet described. Not the kind you could measure in quarters or vesting cliffs or the window between a roadshow and an IPO. The other kind. The kind that belongs to you and you alone.

I didn't write a note. That would have been one more thing manufactured for someone else's consumption. There was no heroic final speech. No redemption, no realization that building something that mattered was still possible, no eleventh-hour pivot back to the original vision. Just a man who had become a product, quietly decommissioning himself. The code stops running. The process terminates. Exit code zero.

Somewhere in a colo on Page Mill Road, a server rack kept humming, serving targeted ads to users who thought they were part of a community. The recommendation algorithm kept optimizing, serving content to 4.7 million profiles that didn't know their creator had finally gone offline. And on Emerson Street, in the apartment above the garage where nothing had ever been manufactured except an empire, the palms held their breath in the dark and waited for a dawn that belonged to someone else.