Introducing Fuz, a zippy stack for autonomy

[ryanatkn]

There's a stack under every app and software tool, whether observed or not, and each tech stack bakes in choices that affect the experiences of end-users, developers, and operators. Fuz is a web-based stack that aims to be efficient, powerful, aligned, and thoughtfully-designed. It supports building servers, static websites, local-first binaries, and open-ended tools, across the lifecycle from dev to ops. It's pre-alpha, not yet ready for production, and sprawling, so terminology and quality may churn, and there's ample opportunity for contributors to help shape the software.

Fuz wants to make its version of the best possible software experience for end-users, developers, and operators. Since 2019 I've been publishing libraries that build on my favorite technologies, with the goal of supporting the generalized needs of many kinds of apps and tools. The libraries try not to be overly-coupled, and most dependencies are optional. Fuz opts for many repos over monorepos, and there's Rust tools on the way to both complement and extend the web tech. Some important pieces have not been announced or released publicly yet in early 2026 -- see fuz.dev for the main site.

Like all stacks, Fuz makes some specific tech choices, so although it can be used a la carte, you'll lose certain benefits or capabilities depending on where you diverge. Sometimes libraries are more generic than this list would imply, for example fuz_css works for JSX/React/Preact/Solid. But generally speaking, Fuz chooses:

• Svelte and SvelteKit
• TypeScript
• plain CSS
• Rust
• PostgreSQL and PGlite for embedded targets (and the Postgres-compatible cloud formations)
• WASM, WASI, component model (for language interop)
• Deno, NodeJS, and Bun

It's early 2026 is when I'm starting to talk about Fuz, after almost 7 years of quietly coding the pieces in public as my full-time funemployment. Fuz's formulation as a stack didn't happen until 2025 -- I had been focused on a white whale of an everything-app while building a collection of modular libraries and tools, but I dropped the "one big app" framing and decided to build many apps with the stack, which had the side effect of validating the years of effort I had put into making projects like fuz_css and fuz_ui generic and reusable. (was it the best use of my time when the industry-standard Tailwind and Shadcn exist? debateable, but I like the results, and they're proudly human-designed, pre-dating my AI usage)

This framing of stack-with-swappable-parts gives users more choices, decentralizes control, and should allow more fluid evolution, as Fuz's parts can adapt to changes and convergences in the wider web/TS/Svelte/Rust ecosystems more easily when they're smaller and with well-factored concerns, and stakeholder groups can form around parts people care about. Instead of being one everything app, Fuz will ship many apps from its many parts! (and look out for fuz_os)

I don't think things are stable enough yet for non-experts to use Fuz, but things are progressing quickly, and the goal is to provide tools to people who don't identify as technical. There's a lot of interconnected projects here and the big picture may not be legible, but I aim to build a stack not a labyrinth! fuz.dev is a fine place to start, and see also the fuz-stack skill docs for agents. Better docs are coming.

Note that @fuzdev has PRs closed to non-maintainers on most of its repos, but I'm soliciting all kinds of community help. For more see Contributing to Fuz.