Why We Burned It Down - John's POV (the chaos)

tl;dr

We rewrote Char because what we built before was confused. The only way to fix it was to start over.

The Early Mess

When Yujong and I started, we were trying to build an AI hardware toy that talks to kids. Hardware was too slow for two people to ship fast, so we pivoted to software.

I'd spent time in meetings using AI notetakers like Clovanote. They missed nuance—the difference between "ok" and "OK!" matters when you're selling. I wanted something simple, clean, fast. Something I'd actually use. That became Char.

Shiny Tech. No Philosophy.

Early on, we got distracted. We picked Tauri because it felt cool. We committed to local-first everything because it sounded clever. Transcription, LLM, cloud, BYOK, lifetime license—we had no product philosophy. No sharp segment. Just vibes.

People liked it initially. Open source folks said they'd use it over Granola just because it was open source. That gave us confidence, but it also confused us. Local-first and open source aren't the same thing. We never picked a lane.

YC Made It Worse

Then Y Combinator happened, and we stopped thinking clearly.

Our group partners kept asking the right questions. The one that stumped us most was:

Who are your users?

We said we were focusing on enterprise, but we weren't sure. We were still deciding between banks, defense, military, or whoever. Instead of slowing down to think, we tried to answer immediately. We confused speed with clarity.

As James from Pioneer Fund told me:

Don't mistake motion for progress.

He said it's written on his son's door. We fell into that trap hard. We were moving fast in 12 different directions at once.

Too Much Logic. Too Much Talking. Too Little Clarity.

We debated everything: local vs cloud, enterprise vs prosumer, bottom-up vs top-down. We tried to logic our way into clarity. You can't.

Then we added hiring, which made it worse. Too many opinions. Too much communication overhead. Alignment took forever. We moved slower, not faster.

YC was a wake-up call on hiring for me. I'm not hiring again for a while. Just me and Yujong now—and we're moving faster than ever.

Investors Pushed One Way. My Gut Said Another.

Early angels pushed the "local-first enterprise" angle. Regulated industries do care about data sovereignty. But enterprise and local-first aren't the same. They care about compliance, SOC2, consent, legal—we weren't ready for any of that.

The whole time, my gut said: Build for people like you. Busy individuals who live in meetings and want a tool that just works. I ignored my gut for months.

The Product Rotted

Because we never committed to a direction, we tried to build everything: local speech-to-text, local LLM, bring-your-own-endpoints, Whisper, an admin server, hybrid setups no one understood. It became an over-engineered pile of potential that felt bad to use. The codebase was tangled. The design was directionless. I didn't even like opening our own app.

Motion Isn't Progress

For months, I was obsessed with motion—metrics, launches, growth hacks. But everything starts and ends with product. I've seen the hype cycles in the valley. They all crash. What's left is the product. Nothing else.

The MVP Isn't What It Used to Be

Around this time, I read something from the Linear blog that clicked.

The definition of "MVP" has changed. In the lean startup era, an MVP could be scrappy—barely working. That's no longer true. Users expect more. Products have matured. An MVP has to feel complete. It can be scrappy under the hood, but not in the hands.

If your wedge is design, like ours, you don't even get that scrappy margin. You have to nail it.

So we stopped playing the "ship crap fast" game. We're building something tight, sharp, and clean—and shipping it with confidence.

Ferrari vs Porsche

I've been thinking about differentiators. Honestly, I don't have a perfect answer. Granola exists. They're great. They're Porsche.

We want to be Ferrari.

Both are sports cars. Both fast. Same league. Ferrari is hand-crafted, opinionated, focused on a specific segment. Porsche is accessible, mass-market.

We're not trying to win everyone. We're trying to dominate a niche that really cares. We'll use a waitlist. We'll charge from the start. If people don't like that, fine. We're not for everyone.

Just Focus

I've stopped caring about what others are building. Competitors matter long term. Right now, it's us, our users, and the product.

Burn It Down. Build It Better.

Fixing the old app would've taken longer than rewriting it. And it didn't deserve to be fixed.

So we burned it down. We rebuilt everything from scratch. New app. Clean architecture. Sharper focus.

This time we're building for people like us. We're opinionated. We're not mistaking speed for progress. We're not shipping crap to look busy. We're going Ferrari mode.

This isn't a polished comeback story. It's us finally admitting what we messed up and getting back to building a product we actually believe in.

Char