Month 64

April was the foundation; May was the part you will actually touch. With the model underneath finally shaped, the work moved up to the interface: the screens, the flows, the moment you go from nothing to a running app. It's the layer I've been wanting to get to for years. It's also the layer that's easiest to get wrong, and the one where going wide is a constant temptation.

What Moved

The new interface stopped being a skeleton and started being a product. A command palette to drive the whole thing from the keyboard, a dashboard that surfaces what needs attention instead of dumping data, environment views for logs and history and settings, a notification center that updates in real time, and an onboarding flow that walks you from an empty account to a deployed app. This is the experience layer I said the days were moving toward back in March — it's here now, and it changes how strackt feels to use.

'Your servers, our problem' used to start after you already had a server. Now strackt can get you the server too. Connect a provider account — Hetzner, DigitalOcean, Vultr, Hostinger, UpCloud, Linode, TransIP — pick a size, and strackt orders it, installs it, and adopts it without you ever opening a console. Each provider boots a little differently; making 'any VPS' actually mean any VPS is unglamorous work, and most of it is invisible when it goes right. Alongside that, Layover: a way to try strackt on a server I lend you, so you can watch the whole thing work before you bring your own.

And the features April was supposed to unlock arrived: a fresh environment for every pull request, cloning a deployment with one click, encrypted backups to storage you own. Last month I said May was for testing — running it, breaking it, finding the edges — and it was. A lot of May was breaking things on purpose and fixing what fell out. That doesn't photograph well, but it's how the edges get found before someone else finds them.

What Didn't

Focus. Building wide is seductive — six providers, a second interface, a backup path, an on-ramp — and each one is defensible on its own. Put together, May was a month of breadth, and breadth has a cost: the new interface exists but isn't the default yet, so for now there are two of everything, and the product is more complex than it was, not less. That's the deal I made knowingly — you add before you remove — but living inside it is less comfortable than writing it down.

And the thing I still haven't done: put it in front of someone. The onboarding now exists to make that easy. I keep building the door and not opening it.

Would I Pay For This Today?

Closer to yes than it's ever been. The stability held through another heavy month — this site, the tooling, my own environments all run on strackt and stayed up while I rebuilt the surface on top of them. What's new is that the experience is starting to feel like the product instead of the scaffolding around it. The asterisk from last month — too complex — is shrinking, but it isn't gone: I'd be paying for the promise and the stability, and trusting that the new interface becomes the only one. Soon, not yet.

Next Month

Make the new interface the default and start taking the old one away — the remove half of 'add then remove'. Less building wide, more finishing what's already wide. And the part I keep deferring: actually inviting someone in. Between Layover and the new onboarding, there's finally a clean path for a stranger to go from nothing to a running app on a server strackt ordered for them.

The launch target is still mid-2026; that's close now. Four years of building the door — July might be the month I open it for early access.