lithair is a memory-first Rust web framework I'm building — one compiled binary serves the frontend, runs the cluster, and is the database, event-sourced, no external services when the working set fits. This post stands on its own. Eight weeks after v0.2, it hit 1.0. And I want to be honest about how that actually feels, because it isn't a story about code.
The tag is cut. v1.0.0. I should be celebrating, and instead I'm looking at the screen thinking there's still stuff left. That it isn't quite finished — even though it's tested end-to-end, soaked in production under real traffic, and survived its own upgrade without losing a single record. Objectively: solid. Subjectively: I don't quite believe it yet.
It took me a while to understand why.
What the distrust is actually rating
It isn't rating lithair. The numbers are there: the measured cluster envelope, backup proven by actually restoring, eight roadmap gates closed one by one. The distrust is rating something else — what it is to put something you care about out there under your name.
There's the small voice: "an AI wrote it anyway." And it's true an AI generated a lot of the lines. But a framework isn't its lines — it's its decisions. Choosing to publish the cluster's envelope instead of claiming infinite scale. Making the admin secure by default. Naming the limits instead of burying them. Deciding, after two years, what to build and — more importantly — what to refuse. The AI typed. The idea, the calls, the standard — those are mine. Typing isn't authorship.
And there's the question of whether it gets seen. But I have to be precise here, because this is where I could misrepresent myself: I didn't build it to break through. I built it for myself — to propose my vision and see it hold. People showing up would please me; it was never the goal, and I'm not waiting on it. Traction is a lottery anyway — timing, luck, network. Pinning two years of work to a draw I don't control would be handing my peace to chance, when the real reward is already here: the vision exists.
What's actually there
So I strip the feeling and look at the thing.
Two years ago, lithair was an idea. One stubborn idea: what if a single binary could serve the site, run the cluster, and be the database? Not three services, not six layers, not an infra to operate. A compiled binary, state in memory when it fits, history in an event store, security close to the runtime.
Today that idea compiles. It runs. It's tested. And it holds this very site on its own — three sites, one binary, secure admin, hot-reload. Website, cluster, database, in the same executable. The all-in-one isn't a line on a slide anymore; it's a process with a PID that replays its log on boot and serves pages. That's not nothing. That's the whole thing.
Honest about what remains
A 1.0 isn't a finish line. It's "the floor is solid enough to build on." Optimizations remain — compute throughput, true zero-downtime on a binary upgrade (today it's a brief graceful restart, which I wrote in the docs instead of pretending otherwise). But the starting base is the right one: a stable API contract, on-disk format unchanged since 0.13, and what's there is measured and documented — limits included. That's what I always meant by "production-ready": not "it does everything," but "the envelope is measured and written down." The load plate, bolted to the wall.
What no one can take
The AI codes at the start. Then the rest — the standards, the judgment, the ecosystem built around it, the two years. lithair is just one exhibit of that, next to my other projects. And the idea from two years ago — an all-in-one that holds — is here. Really here.
Maybe I don't quite believe it yet. But it exists now. And that, no one can take from me.
Top comments (0)