[San Francisco Apartment — January 2014, Morning]
The pricing page loaded, and something in Ethan's chest tightened.
V100-equivalent instances: $50 per hour. A100-equivalent: $120 per hour. H100-class: $300 per hour. Minimum billing increments of one hour. No free tier. No startup credits. No "contact sales for enterprise pricing."
Just a text box at the top of the page, rendered in a font that shouldn't exist yet:
You know what future hardware is worth. Pay accordingly.
Ethan scrolled. The interface was clean — too clean for 2014, when most cloud dashboards looked like Excel sheets designed by committee. Instance types, regions (listed as temporal coordinates rather than geographic locations), storage options, networking. Everything a modern cloud provider would offer, except nothing about this provider was modern. Or current. Or possible.
He opened a cost calculator and punched in numbers. A basic Transformer training run — the architecture crystallized in his head — would need at minimum a few hundred GPU-hours on V100-equivalents. Call it three hundred hours at the low end, assuming efficient implementation and a small model. That was fifteen thousand dollars.
His stomach dropped.
He didn't have fifteen thousand dollars. He didn't know if the body he occupied had fifteen thousand dollars. He didn't know what the body had at all.
Ethan pushed back from the desk and looked at the apartment with new eyes. Not the eyes of a man processing impossible circumstances, but the eyes of someone taking inventory.
The closet first. He pulled open the sliding door. Five shirts — two button-downs, three tech company t-shirts. Three pairs of jeans. One pair of dress shoes still in their box. A North Face jacket with a Techstars pin on the lapel. The clothes told a story: someone who dressed for pitch meetings but lived in hoodies.
The desk drawers. A California driver's license: Ethan Gardner, DOB August 12, 1988. Twenty-five years old. An address matching this apartment. A Bank of America debit card. A Chase credit card. A business card — Gardner Analytics, CEO & Founder. Below the name, in smaller text: Data-Driven Solutions for the Modern Enterprise.
He almost laughed. Gardner Analytics. A nothing name for what was almost certainly a nothing company. The kind of venture a first-time founder creates when they know the buzzwords but not the business.
More drawers. A lease agreement — month-to-month, $1,800. Utility bills. A letter from the California Secretary of State regarding Gardner Analytics Inc., filed seven months ago. Bank statements.
The bank statements told the real story.
Savings account: $12,341.16. Checking: $2,100.87. The savings balance had been falling for months — $18,000 in September, $16,000 in October, $14,000 in November, $12,000 in December. Consistent withdrawals. Rent, food, AWS bills for a product nobody was buying. The trajectory was a straight line aimed at zero, arriving sometime in February.
Two months of runway. Maybe ten weeks if he stopped eating out entirely.
Ethan sat on the bed with the bank statements fanned across the comforter. Gardner Analytics had one client — a small marketing firm paying $500 a month for a dashboard that probably ran on a single EC2 instance. The company's GitHub had twelve repositories, all private, all last committed to in October. The founder had given up coding before he gave up the company.
No co-founder. No employees. No investors. No advisors. The LinkedIn profile had 340 connections, mostly other failed founders and a handful of recruiters.
Nobody would miss the previous Ethan Gardner. The realization sat heavy. A man had existed here — eaten at this counter, slept in this bed, dreamed startup dreams in this apartment — and the universe had swapped him out like a defective part. No one had called in days. The voicemail was empty. The text messages were spam and one-sided conversations with a college friend who'd stopped replying in November.
The cereal was in the cabinet above the sink. Cheerios. Generic brand. Ethan poured a bowl without milk — there was none — and ate standing at the counter, each dry handful tasting like cardboard and guilt. Another man's cereal. Another man's spoon. Another man's life, and all of it circling the drain.
He chewed. Swallowed. Poured another handful.
Focus.
The laptop pulled him back. He opened a code editor — Sublime Text, the 2014 favorite — and created a new file. transformer.py. His fingers hovered over the keyboard.
The architecture in his mind pulsed. Clear. Waiting.
He started typing. Import statements first. NumPy. Theano — because PyTorch didn't exist yet, TensorFlow was still a Google internal project, and Theano was the best framework available in January 2014. His fingers moved, and something shifted.
The code came fast. Not the normal kind of fast, where you type and think and delete and retype. This was different. His hands knew where to go before his conscious mind finished the thought. Import declarations flowed into class definitions, into matrix operations, into the attention mechanism itself — the scaled dot-product attention that was the beating heart of the Transformer.
Two hundred lines of CUDA kernel code in twenty minutes. Clean. Commented. Structured the way a senior ML engineer would write it, not the way a mid-level practitioner would fumble through it.
Ethan stopped typing and stared at the screen. His hands tingled. The code was correct — he could feel it the way a musician feels a chord progression resolving. But the speed was wrong. No human typed production-quality CUDA that fast. Especially not CUDA for operations that, in this timeline, hadn't been theorized yet.
Another ability. Like the architecture in his head, like ChronoCloud in the browser — something given, not earned. His coding speed for ML-specific work wasn't just enhanced. It was supernatural.
He saved the file. Opened a terminal. Tried to compile.
Errors. Of course. Theano's CUDA interface in 2014 didn't support half the operations he'd written. The code was correct for 2025 tooling. Useless for what existed now.
The gap between knowing and executing. The architecture was in his head. The coding ability was in his hands. But the tools of 2014 couldn't bridge the distance. He'd need to adapt — write custom kernels, build compatibility layers, fight the framework every step of the way.
And he'd need compute. Real compute. ChronoCloud compute. Which meant money he didn't have.
Ethan closed the laptop. Crossed to the bathroom. Ran the faucet and splashed water on his face.
The mirror showed a stranger. Thinner jaw than his original face. Brown hair instead of black, slightly wavy. A nose that had been broken at least once. Brown eyes — his original eyes had been green. The overall effect was someone unremarkable. Forgettable. The kind of face VCs looked past in a pitch meeting.
Different face. Same mind.
He gripped the sink edge and stared at the reflection until it became familiar. Or at least less alarming.
Pied Piper. The name surfaced like a bubble. Richard Hendricks. The awkward genius with the compression algorithm. What else? Erlich Bachman — the loud one, the incubator guy, Aviato. Gilfoyle and Dinesh, the two coders who traded insults like currency. Gavin Belson, the Hooli CEO with the god complex. Jared, the business guy with the traumatic backstory. "This guy fucks" — that was about someone. Russ Hanneman. The billionaire with the car doors.
Then the edges blurred. Season three. Season four. Something about a video chat app? A blood boy? Jian-Yang committing some kind of fraud? The details dissolved like sugar in water. He'd binged the show during COVID lockdown, six seasons in a week, half-watching while training models on his second monitor. Entertainment, not study material.
His meta-knowledge was a shallow pool. Wide but ankle-deep.
Ethan dried his face. Hung the towel on the rack. Walked back to the desk.
TechCrunch Disrupt. The startup competition. Season one, episode one — or close to it. Richard Hendricks unveiling Pied Piper's compression algorithm. The Weissman Score that broke everything. That was coming. Soon. Within weeks, if the show's timeline mapped to the calendar on the wall.
A room full of Silicon Valley's most important people, all gathered in one place. Founders, VCs, press. The starting line of a story he half-remembered.
A place to begin.
---
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