Winding Down Kreye

Some projects end because they failed. Some end because they succeeded and you sold them. And some — most, honestly — end somewhere in the quiet middle: they worked, people used them, you learned a lot, and one day you realize you’re keeping the lights on out of sentiment rather than conviction.

Kreye is in that third category. Today I’m shutting it down.

What Kreye was

Kreye was an AI-powered dynamic workspace — a canvas-based productivity tool that generated UI widgets from prompts. You’d describe what you wanted, and the app would assemble rich text, tables, checklists, charts, and previews onto an infinite canvas you could drag, resize, and link together. The idea was a workspace that thinks with you: scattered thoughts in, structured insights out.

Under the hood it was two repos. A Next.js front end on a ReactFlow canvas with a JSON-schema-driven widget renderer, and a FastAPI backend doing the AI orchestration, auth, billing, and storage. It was a genuinely fun system to build — the kind of project where the architecture itself was half the reward.

Why I’m shutting it down

Two reasons, and I’ll be blunt about both.

The first is activity. Kreye had users, and I’m grateful for every one of them, but the curve never bent the way a product needs it to. People would try it, build a canvas or two, and drift off. That’s useful signal, not a tragedy — it tells you the thing you made is interesting but not yet necessary.

The second is cost. An AI workspace is not cheap to run. Model calls, storage, a database that’s always on, the long tail of infrastructure that quietly bills you whether ten people log in or zero. For a product carrying low activity, every month became a small donation to my own cloud provider. At some point the spreadsheet stops being ambiguous.

Put those together and the decision makes itself: I was paying real money to keep a service running that I was no longer actively building toward. The opportunity cost — my time and attention — was the expensive part.

Where my attention is going instead

The flip side of winding something down is honesty about what you’d rather be doing. I have three projects I’m genuinely more excited about right now, and they deserve the focus Kreye was diluting:

• Tandemu — measuring how developers actually work with AI coding assistants. AI-to-manual code ratios, friction detection, commit-level attribution, and a persistent memory layer for your AI teammate. This is the one I keep wanting to think about on walks.
• AnswerSEO — SEO for the era of answer engines. Getting cited by LLMs and AI search instead of fighting for blue links.
• Foodbang — a more product-and-people project, and a nice counterweight to the heavily-AI rest of my stack.

None of these are hypothetical pivots. They’re already where my best hours go, and Kreye was the thing I had to consciously not work on to make room for them. Shutting it down just makes the calendar match reality.

Open-sourcing the code

Here’s the part I’m actually happy about: Kreye isn’t disappearing into a private archive. I’ve cleaned up both repositories — scrubbed the history of secrets, parameterized the config — and I’m putting them out in the open:

• Backend (FastAPI): github.com/sebastiangrebe/kreye
• Frontend (Next.js): github.com/sebastiangrebe/kreye-frontend

If you’re building anything in the AI-workspace space, there’s a lot in here worth reading: a schema-driven widget rendering system, a ReactFlow canvas with semantic linking, a clean React Query + Zustand data layer, and a FastAPI backend with auth, Stripe billing, and AI orchestration already wired up. Fork it, lift pieces out of it, or just poke around the architecture. That’s a far better fate for the code than a dormant production deployment.

A thank you

If you used Kreye — signed up, built a canvas, sent me a bug report, or just clicked around for ten minutes — thank you, sincerely. You gave a side project the one thing it can’t generate for itself: real people doing real work in it. I learned an enormous amount from watching that, and a good chunk of what I learned is going straight into the projects above.

Building things is easy to start and hard to stop. Stopping well — saying thanks, leaving the code open, and pointing your energy at what matters more — feels like the right way to close this one out.

On to the next.

If you want to follow where I’m headed, the projects page is the live list, and you can always reach out.