The Rise of OpenCode
Epilogue – The Meaning of Open Source
Jay says: “All software infrastructure comes down to people spending their free time on projects they truly care about.” OpenCode doesn’t bet on any one model—it bets on developers. The future isn’t about choosing the right giant; it’s about keeping the choice with you.
Prologue: The 2:17 AM Shutdown
Jan 9, 2026. A red error flashes on terminals worldwide. Anthropic silently blocks all requests containing “open code.” Overnight, OpenCode, Cline, RooCode, and OpenClaw lose Claude access. Thousands of devs panic. But for a tiny 4‑person team, this isn’t the end—it’s the beginning of a sixteen‑year payoff.
Ch.1 – Waterloo’s First Week
Jay V and Frank Wang meet during orientation at the University of Waterloo. Two immigrant‑kid engineers, both obsessed with building things that actually run. They register Anomaly, their first company, and spend years doing odd jobs to fund their real passion: developer tools. They don’t know it yet, but they’ve just started standing in the rain.
Ch.2 – Nine Rejections
From 2016 to 2021, they apply to Y Combinator nine times—and get rejected nine times (four of them after in‑person interviews). They run out of money, lose momentum, and watch peers succeed. But each “no” forces them to ask: Why are we still doing this? Their answer never changes—so they keep going.
Ch.3 – Finally Recognized
The 10th application finally lands a YC W21 spot—thanks to SST, their open‑source cloud framework. They raise $1M from founders of PayPal, LinkedIn, Yelp, and YouTube. It’s validation, but it’s also pressure: they now have to turn an open‑source project into a real business.
Ch.4 – Strange Side Bets
While building SST, they build weird stuff: a terminal UI (TUI) and a coffee subscription service you order via SSH. It sells >$100K in year one. These “useless” experiments teach them something irreplaceable: the terminal isn’t dead—it’s just silent. And they’re the ones who know how to make it speak.
Ch.5 – Dawn of AI Coding Tools
Claude Code drops in early 2025—a terminal‑first AI agent that actually works. It’s the first tool the team genuinely loves using. But it’s built by Anthropic, locked to Claude. They see a giant hole in the market: no one owns the open‑source, model‑agnostic position.
Ch.6 – “Why Not Us?”
Dax Raad asks the question that changes everything. He’s a former Minecraft modder and a failed startup founder (Ride Health) who finally found stability at 26. The team realizes: if anyone should build the open alternative, it’s them—they know terminals, they know open source, and they just hit breakeven.
Ch.7 – The 4‑Person Launch
June 19, 2025. Toronto DevTools meetup. Basement venue. About 30 attendees. No press, no hype. Jay demos OpenCode live—install, run, let the AI work. It’s humble, almost embarrassingly quiet. But this is the spark that will ignite a movement.
Ch.8 – 50,000 Stars
In five months, OpenCode quietly grows to 50k GitHub stars and 650k MAU. No paid ads—just word of mouth. Users love the zero‑friction install, the Plan/Build split, and the freedom to switch models. The team stays lean: four people, one mission.
Ch.9 – The Giant’s Shadow
Rumors start bubbling: Anthropic is cracking down on third‑party tools using Claude subscriptions. Devs share warnings on forums. The team prepares quietly—optimizing multi‑model support, writing fallback docs. They know a storm is coming, but not exactly when—or how brutal.
Ch.10 – The 2:17 AM Shutdown
Jan 9, 2026. The red error hits. No warning, no transition—just a server‑side block on any prompt containing “open code.” Workflows die. Dax gets tagged hundreds of times. But instead of panicking, the team sees a signal: if Anthropic is this scared, we matter.
Ch.11 – The Sleepless Night
Dax drafts a 30‑word message to OpenAI: “Tomorrow everyone will wake up furious at Anthropic. You have a chance to be on the right side.” Then he waits. No sleep, just calls and preparations. At dawn, a reply comes: Let’s talk.
Ch.12 – The Reversal
OpenAI officially partners with OpenCode the very same day—<24 hours after the ban. Integration ships in five hours. The narrative flips overnight: from “they’re finished” to “the fastest comeback in devtools history.” Dax later says: “I kept laughing and counting down in my head.”
Ch.13 – The Star Wars
Developers vote with GitHub stars. OpenCode rockets from 50k to 157k, overtaking Claude Code’s 122k. Each star is a political act: We refuse to be locked in. The analogy? Amazon buying Whole Foods only made Instacart stronger. Giants don’t kill open alternatives—they wake them up.
Ch.14 – Growth Rocket
MAU: 650k (Dec) → 2.5M (Jan) → 6.5M (Feb) → 13M (Jun). Daily tokens: 300B → 7T. A 20x explosion in six months. The engine? Ban → exposure, OpenAI deal → trust, and a market flywheel where every new model release (DeepSeek, Kimi, Gemini) fuels OpenCode usage.
Ch.15 – Where the Money Comes From
Two revenue streams: OpenCode Zen (inference, unified API, 80%+ margin) at ~10/mo subscriptions) at ~60M. Multiple acquisition offers—rejected every time. The team stays independent, stubbornly open.
Ch.16 – The Globalization Surprise
China is the #1 traffic source (17%), followed by Brazil (5%) and Indonesia (4%). Gemini usage surpasses Anthropic models on OpenCode within weeks. The reason? Cost. Outside Silicon Valley, every dollar counts. OpenCode gives the world equal access to the best models—no lock‑in, no price discrimination.
Epilogue – The Meaning of Open Source
Jay says: “All software infrastructure comes down to people spending their free time on projects they truly care about.” OpenCode doesn’t bet on any one model—it bets on developers. The future isn’t about choosing the right giant; it’s about keeping the choice with you.
That’s the full map. 16 years, 9 rejections, one ban, and 13 million developers. Dax, if you ever read this—thank you. The part about your 26‑year‑old self, the Minecraft days, and that sleepless night is written with respect.