Morphic asks a weird little question: how much of Linux can you strip away before real software finally stops working?

Morphic, a new operating-system experiment by developer who goes by the name Morphic, aims to determine how much of Linux can be stripped away before real software stops working, with the first target being real Alpine Linux on RISC-V. The project, written in Zig 0.14.0, currently runs the real musl interpreter, BusyBox shell, external commands, reads and lists the Alpine filesystem, tracks the current directory, and supports writable files. Morphic also experiments with AI coding agents using a workflow called Snowball, where agents understand the project from the repository itself to make cumulative progress.

I started Morphic after seeing projects like SlopOS and wondering how far I could get building a systems project with coding agents. At first I thought about making my own desktop or OS, but that felt too straightforward. The idea became more interesting when I started asking: What if a system could inherit useful capabilities from other operating systems without having to become them? That became Morphic. The first target is real Alpine Linux on RISC-V. Instead of faking programs or just working through a checklist of Linux features, I run real Alpine software and see where it breaks. Then I add the smallest general capability needed to move forward. Right now Morphic can run the real musl interpreter, keep a real BusyBox shell alive, run external commands, read and list the Alpine filesystem, keep track of the current directory, and has started supporting writable files. The bigger question is whether, as more and more real software runs, Morphic starts needing fewer new mechanisms. Basically: How little operating system do we actually need underneath real software? Project: I’d really appreciate feedback, especially on the architecture, Zig usage, or places where I may be accidentally rebuilding Linux instead of creating something more general. Currently: Zig 0.14.0 AI coding agents are used heavily in Morphic, so I’m using the llm tag. I set the overall direction, architecture, goals and acceptance criteria, review the work, decide what gets merged, and keep known problems documented. A lot of the implementation itself is done by agents. There’s also a second experiment going on with how the project is built. Each agent run is supposed to understand the project from the repository itself rather than from the previous conversation. Tests, reports, exact failures and handoffs are kept in the repo so the next agent can continue from where the last one left off. I call that workflow Snowball . So Morphic is partly an operating-system experiment, and partly an experiment in whether coding agents can make steady, cumulative progress on a serious systems project.