[Meridian Ventures, Sand Hill Road — Early February 2014, 2:00 PM]
The second VC's office had a fountain in the lobby. Not a wall fountain or a tabletop model — a full-sized, free-standing fountain with three tiers and a recirculating pump that filled the reception area with the sound of running water. It was supposed to convey serenity. It conveyed that someone had spent ten thousand dollars on lobby furniture.
Ethan's shoes were scuffed. He'd noticed this in the parking lot and rubbed them on the back of his calves, which accomplished nothing except transferring the scuff to his pants. The messenger bag's strap had frayed another quarter-inch since the last pitch. His button-down was clean but wrinkled at the collar where the iron — the dead man's iron, found under the bathroom sink — had failed to reach.
The founding partner of Meridian Ventures was a woman named Patricia Liang. Mid-forties, Stanford MBA, former Google engineer. David Park had made the introduction via email, a chain that included the line "Patricia has a technical background — she might actually get what you're building."
The emphasis on "actually" had been both encouraging and insulting.
Patricia met him in a conference room with glass walls and a whiteboard that took up the entire north side. She shook his hand with the grip of someone who'd started her career writing C++ and transitioned to writing checks. Her first words were: "David says you're doing something interesting with neural networks. I spent three years on the Google Brain team before I went into venture. Don't dumb it down."
Permission to be technical. A gift.
Ethan opened the laptop and went technical. Not the consumer-friendly nine slides — a new deck, built with Sarah over the past three days. This version had architecture diagrams. Attention mechanism schematics. Training curves from the test model. A slide showing the generated coffee shop paragraph with annotations explaining which architectural features produced which qualities in the output.
Patricia followed. She followed. Her questions were specific: "What's the computational complexity of scaled dot-product attention relative to sequence length?" and "How does positional encoding handle variable-length input?" and "What's your strategy for preventing mode collapse during generation?"
For twenty minutes, Ethan was pitching to someone who spoke his language. The relief was physical — a loosening in his shoulders, a steadiness in his voice that hadn't been present at Basecamp or the bar afterward. Patricia understood self-attention. She understood why transforming all positions simultaneously was more powerful than processing them sequentially. She understood the implications.
Then the conversation turned.
"The technology is fascinating," Patricia said. She set down her pen — she'd been taking notes in a leather-bound journal, another signal that she took this seriously. "But here's my problem. I invest in companies, not research projects. Who buys this?"
"Content generation. Marketing, legal, customer service—"
"Those markets exist, but they're served by humans. Humans who are cheap, reliable, and don't hallucinate. Your model — even at full scale — will produce text that's good but imperfect. Who pays for good but imperfect when they can pay a freelancer for perfect?"
"The scale argument. A freelancer writes one document. The model writes a thousand. The cost per unit drops by orders of magnitude—"
"Assuming quality. Assuming consistency. Assuming liability protection — if your model generates something defamatory, who's responsible? The company that deployed it? The company that built it?" Patricia leaned back. "I spent enough time at Google to know that language generation is extraordinarily hard. What you're showing me is impressive for a two-person team. It's not impressive enough for a market that doesn't exist yet."
"The market will exist."
"When?"
"Three to five years."
Patricia's expression shifted. Not dismissal — sympathy. The look of someone who knew the founder was probably right but couldn't justify the timeline to her LPs.
"Ethan, I like the technology. Genuinely. If you'd come to me in 2018, I'd write this check in a heartbeat. But my fund has a five-year return horizon. I can't put money into a market that's three years from materializing and hope it returns within the window. The math doesn't work."
The math doesn't work. His own phrase, reflected back at him by someone who meant it in the opposite direction.
"Is there anything I could show you that would change your mind?"
Patricia considered. "Paying customers. If you can find even one business willing to pay for generated text — not a pilot, not a beta, but actual recurring revenue — that changes the conversation. Revenue means market. Market means timeline. Timeline means I can model the return."
She stood. Extended her hand. "Keep going. Seriously. This technology has a future. I just can't fund the future three years early."
---
[Parking Lot, Meridian Ventures — 2:35 PM]
The Honda Civic's driver seat had a spring that poked through the upholstery if you sat at the wrong angle. Ethan found the angle. The spring jabbed his lower back like an accusation.
He didn't drive. Not yet. He sat in the car with the engine off and the windows up, breathing the stale air of a vehicle that had been parked in a garage for months before he'd retrieved it.
Two pitches. Two rejections. Different reasons, same result. Alan Rao couldn't understand the technology. Patricia Liang understood it perfectly and couldn't justify the timing. The first was a communication problem. The second was a reality problem. And reality was harder to solve.
He thought about the bar. The Jameson. The bartender who'd asked if he was building Siri. He didn't want a drink today. The failure was different — cleaner, more honest. Patricia had treated him like a peer. She'd engaged with the architecture, asked questions that demonstrated genuine comprehension, and given him a rejection grounded in rational analysis rather than confusion.
Somehow that was worse. Confusion could be solved with a better pitch. Rationality required a different kind of answer.
His phone buzzed. Sarah.
How'd it go?
Ethan stared at the text. The cracked screen caught a shard of afternoon sunlight through the windshield.
She understood the tech. Didn't believe the market.
Three dots. Sarah typing.
Then we need to make the market believe.
A pause. Then another message.
Also — you got a reply. From Raviga. Check your email.
Ethan opened Gmail. The inbox loaded slowly on the 5S's aging processor. Twelve new emails. Eleven spam. One from [email protected].
Mr. Gardner — I remember you from Disrupt. You were the one in the back row who wasn't checking his phone during the compression demo. I'd be happy to meet. I'm at a Bay Area Startup Alliance event next Thursday. We could talk there, or I can do coffee the following week. — Monica Hall
He read it twice. Then a third time. She remembered him. Not from a cold email — from Disrupt itself. She'd noticed him noticing things, the same way he'd noticed her.
The spring in the driver's seat was still jabbing his back. He shifted. It didn't help.
Two rejections. One open door.
Ethan started the car.
Author's Note / Promotion: Your Reviews and Power Stones are the best way to show support. They help me know what you're enjoying and bring in new readers! You don't have to. Get instant access to more content by supporting me on Patreon. I have three options so you can pick how far ahead you want to be: 🪙 Silver Tier ($6): Read 10 chapters ahead of the public site. 👑 Gold Tier ($9): Get 15-20 chapters ahead of the public site. 💎 Platinum Tier ($15): The ultimate experience. Get new chapters the second I finish them . No waiting for weekly drops, just pure, instant access. Your support helps me write more . 👉 Find it all at patreon.com/fanficwriter1