You Don't Get to Take It Offline

There's an easy way to rebuild a product.

You take it down. You throw up a "we'll be back soon" page. You rebuild in peace, flip a switch, and unveil the new thing to applause.

We didn't get that version.

The app we're rebuilding has members on it. Right now. While I'm typing this, someone is logging in, posting, watching, paying — on the old thing we're in the middle of replacing.

You don't get to take it offline.

That one fact changes everything about how you build.

Chapter One: Bubble Did Its Job. I Won't Pretend Otherwise.

Nothing Held Back started on Bubble. No-code. It still lives at v1.nothingheldback.com — still running, still serving members, while we rebuild the whole thing underneath them.

Here's the line people expect from an engineer: that no-code was a mistake. That real developers don't touch it. That we should be a little embarrassed by where this started.

I'm not.

Bubble put a real product in front of real people, fast. It made money. It built a community. It proved the thing was worth building — which is the only question that matters at the start.

You don't apologize for the tool that carried you to the point where you outgrew it.

You just don't stay one day longer than it serves you.

Chapter Two: The Ceiling You Feel Before You See It.

No-code has a ceiling. You don't hit it on day one. You hit it the day the product gets ambitious.

You want a recommendation engine that actually understands what a member cares about. You want data you own, queries you control, performance you can tune down to the index. You want to build the thing you can see in your head — not the thing the platform allows.

Bubble trades control for speed. Early on, that's a great trade.

Then one day the thing you want to build sits on the other side of a wall the platform won't let you cross.

That's the day you stop patching and start rebuilding.

Chapter Three: Rebuilding the Plane While It's in the Air.

This isn't a sandbox. NHB 2.0 is in early production — invite-only at the door, but live members use it every single day.

So the team treats every change as exactly what it is: a real production change, on a real product, that people pay for.

We're moving 29 modules — forum, chat, library, billing, coaching, all of it — one at a time, and the member is never supposed to feel the seam. The old app keeps running. The new one comes up beside it. Nobody gets a "we'll be back soon" banner.

That's the discipline nobody warns you about.

Rebuilding is easy when nobody's watching. Rebuilding while your members are watching — without dropping a single one of them — is the actual work.

You don't get to break things. You get to replace them, quietly, while they're still in someone's hands.

Chapter Four: New Bones. And the Layer I Get to Build.

Here's what NHB runs on now.

Next.js 16 up front. Supabase and Postgres underneath. Drizzle for the schema, hand-written migrations because we want to know exactly what touches the database. Tailwind for the UI. Every layer picked on purpose — not because a platform picked it for us.

But the stack isn't the story. This is.

My lane is the layer Bubble was never going to let us build. An AI recommendation engine. Embeddings that actually understand the content. A ranking engine that decides what each member sees next. The rails that carry it across the app.

That's not a feature you bolt onto no-code.

That's the whole reason a team walks away from it.

And Claude is my co-engineer through all of it — same as it's been since day one. It doesn't write the vision. It writes the code at the speed of the vision.

Why the Rebuild Is the Real Work.

The market didn't ask us to rebuild NHB. It told us — the same way it's told me everything.

It told us when the product got bigger than the tool. When "good enough" started costing more than a rebuild would. When the thing we wanted to build stopped fitting inside the thing it was built on.

So we moved.

Not by burning down what worked. By replacing it, live, without the people who trust it ever feeling the ground move under them.

That's the version of rebuilding nobody puts in the highlight reel. No relaunch party. No "we're back" post. Just members opening the app one day to something faster, sharper, smarter — and never knowing how much shifted beneath them.

That's the job.

You don't get to take it offline. So you get good at rebuilding while it's still running.