> I assume this project is all vibe-coded?

You know what happens when you assume, right?

https://github.com/duxlabs-dev/dux-shell?tab=contributing-ov...

Quoting in part from that:

> Using an LLM to help write code is fine. Submitting its output unread is not.

> * You are responsible for every line you submit. You have read it, tested it, and can explain it in review.

> * Say in the PR when a meaningful part of the code was AI generated.

> * Write PR descriptions, issues, and comments yourself. No AI-written or AI-rewritten text. A few plain sentences in your own words beat a page of generated summary. PRs and issues with AI-written text get closed without review.

> ...

> * PRs that read like unreviewed output, with narrating comments, invented APIs, or style that ignores the file they are in, get closed without review.

That looks very clearly non-vibe coded to me. (Also, it's a mature project with 79 releases over the last year and a bit, and 321 contributors.)

As much as I respect the abilities of agentic coding LLMs, I don't think we're yet at the point that you could vibe-code a novel Wayland DE. This is not trivial stuff!

There’s a new vibe coded Wayland shell every week now. It’s actually unfortunate DMS renamed when they did. I saw this and assumed it was just another vibe coded abandonware without realizing what it was.

It isn't really a wayland DE. It's a shell that can be used with niri, hyprland, sway, etc. And is also built on top of quickshell.

Glad to be proven wrong. :)

> You know what happens when you assume, right?

Judging by the text you quoted, you guess correctly?

> As much as I respect the abilities of agentic coding LLMs, I don't think we're yet at the point that you could vibe-code a novel Wayland DE.

You definitely could. Have you tried?