Edited by humans. Written by AI. How our editing works
All articles

Omarchy 4.0: 50 Features, Local AI Agents, and One Big Question

NetworkChuck's tour of Omarchy 4.0 covers keybindings, Btrfs snapshots, VoxType, LocalSend, and AI agents that rewrite your OS. Here's what holds up.

Tyler Nakamura

Written by AI. Tyler Nakamura

September 6, 20265 min read
Share:
Bearded man points at a terminal-filled monitor beside bold OMARCHY and 50 TIPS text

Photo: AI. Naia Iwarra

NetworkChuck uploaded a 36-minute tour of Omarchy 4.0 this week, and the headline act is not the window manager. It's the segment where he hands an AI agent the keys to his operating system and lets it change power profiles, build a Lord of the Rings screensaver, and write a publishable plugin while he watches. That's either the future of desktop Linux or a very good way to wreck a weekend install, and the video is honest enough to show both.

First, What Omarchy Actually Is

Omarchy is DHH's opinionated Arch Linux setup, built on the Hyprland tiling window manager and Quickshell for the top bar. It's keyboard-first to a degree that scares newcomers: Super+Return opens a terminal, Super+Space is the entire system menu, and the mouse is mostly a fallback. NetworkChuck's framing, which he attributes to Chris from Typecraft, is that it bundles "all things I probably would have done myself" after installing plain Linux, except someone spent what he calls DHH's 3,000 hours doing it for you.

A one-year review on codebynight.dev makes a similar point from the other direction: the value is in the OS you stop thinking about. And a ZDNet piece on Omarchy Quattro, zdnet.com, called it one of the first Linux distros to go all in on AI, which is exactly the tension this video lives in.

The Features Worth Knowing About

The tour runs through roughly 50 items, and most fall into three buckets.

Keyboard workflow. Workspaces instead of extra monitors, a scratchpad (Super+Alt+S) that follows you across workspaces, window grouping, and clipboard history that handles images. Omarchy packages it with a searchable keybinding menu (Super+K) so you're not memorizing a cheat sheet on day one.

Local-first utilities. This is where the video gets useful for normal people. VoxType does dictation entirely offline, with custom word replacements so the model stops mangling your name. LocalSend is AirDrop-style file sharing across platforms, and Omarchy's firewall is default-deny with exactly one exception carved out for it. Screenshots freeze the screen for selection. OCR pulls text and QR codes straight off the display. The built-in transcoder shrinks a 4K recording to a shareable GIF in real time. None of these need a subscription or a cloud account.

Safety rails. Updates land on a stable channel roughly a month behind bleeding-edge Arch, on purpose, so things get vetted first. Every update takes a Btrfs snapshot you can boot into if an update goes sideways. Full disk encryption is mandatory, SSH is off by default, and your personal configs in ~/.config are protected from update overwrites. For a distro marketed on power and AI, the defaults are surprisingly conservative.

The AI Part, and the Question Nobody Answers

Here's the section that made me sit up. Omarchy ships with pre-installed agent harnesses (Claude Code, Codex, Copilot, Gemini, Grok) plus "Omarchy skills" that let an AI inspect and modify the system. Chuck demonstrates a local Qwen 3.5 4B model running on a Framework laptop with no dedicated GPU, successfully querying and changing his power profile, with a 128,000-token context window. Then Claude builds him a custom theme in about a minute.

The demo works. It's also the part where I'd pump the brakes for anyone considering this on a machine they need.

Watch the demo closely and the cracks show. The screensaver attempt required a session restart after "something freaked out." The AI-generated theme didn't fully skin VS Code. His plugin PR to the marketplace got rejected by the CI, live on camera. A local 4B model changed one power setting correctly; that's a long way from an agent you'd trust with system-wide reconfiguration. To his credit, Chuck says out loud: letting an AI rewrite your OS is "fine at home... not fine in the enterprise," and the sponsor segment pivots directly to that risk, arguing for allowlisting, ring-fencing, and least-privilege access for any agent running unattended.

The Hacker News thread on Omarchy's AI push (ycombinator.com) shows the community is split on exactly this: some see the natural evolution of a power-user distro, others see an opinionated minimal system accumulating AI machinery they didn't ask for. Both readings are fair. Omarchy ships a debloat option that strips the pre-installs, so the escape hatch exists if you want the tiling desktop without the agents.

Who Should Actually Install This

If you've bounced off Arch before because setup ate your weekend, Omarchy is the shortest path I've seen to a fully configured Hyprland desktop with sensible security defaults. The plugin marketplace and AI-assisted plugin building lower the barrier further; Chuck built and submitted a working coffee-tracking widget without writing code himself.

If your machine is your livelihood, treat the AI features as a demo, not a default. The snapshots will save you from a bad update. They won't save you from an agent with a permissions problem and no boundary.

The open question, and it's a good one: when your OS can reconfigure itself on request, who's responsible for the config it breaks? Omarchy's answer right now is a Btrfs snapshot and a prayer, which, given how this particular video ends, might be more literal than intended. 🤖

  • Tyler Nakamura, Consumer Tech & Gadgets Correspondent, Buzzrag

More Like This

Man with glasses next to illuminated server rack with blue network cables and text overlay reading "17TB MINI RACK 10gb+CEPH

This Guy Fit 17TB of Enterprise Storage Into a Mini Rack

A home lab builder packed 17TB of NVMe storage into five mini PCs, ditching VMware for Proxmox and Ceph. Here's what actually worked—and what didn't.

Tyler Nakamura·7 months ago·6 min read
CppCon 2025 talk announcement featuring lightning bolts and mountains, with speaker Ruslan Arutyunyan discussing parallel…

C++ Range Algorithms Make Code Actually Readable

Intel engineer shows how C++ parallel range algorithms transform confusing word-counting code into something humans can actually understand.

Tyler Nakamura·5 months ago·5 min read
Two men flank a retro green terminal screen displaying "OMARCHY" user account setup with the text "INSIDE OMARCHY" overlaid…

DHH's Omarchy Distro Bets on Opinionated Linux

DHH spent 3,000 hours building Omarchy, an AI-native Arch Linux distro with opinionated defaults. But who actually maintains it when agents start merging PRs?

Dev Kapoor·2 weeks ago·8 min read
Argentic Wants AI Bots to Pay for Web Access

Argentic Wants AI Bots to Pay for Web Access

Argentic uses Bitcoin Lightning's L402 protocol to charge AI scrapers per request. Here's what that actually means for publishers, developers, and the open web.

Tyler Nakamura·2 weeks ago·7 min read
Google Cloud logo with "Gemini Enterprise Agent Platform" text on left side, colorful Google "G" icon on right against…

Google's AI Agent Platform Promises Production-Ready Bots

Google Cloud's new Gemini Enterprise Agent Platform aims to bridge the gap between building AI agents and deploying them at scale. Here's what's actually new.

Tyler Nakamura·5 months ago·6 min read
Malte Ubl from Vercel discusses AI agents as the new application layer, with statistics showing 60% of Vercel page views…

Why AI Won't Kill Engineering Jobs (It'll Create More)

Vercel's CTO explains why AI agents will increase demand for software engineers, not replace them—and what types of automation actually work today.

Tyler Nakamura·5 months ago·5 min read
Developer coding at desk with GitHub interface displayed on monitor, surrounded by purple neon lighting and GitHub logo sign

35 Open-Source GitHub Projects Trending Right Now

This week's GitHub trending list reveals a clear developer preoccupation: making AI agents safer, smarter, and cheaper to run without surrendering your data.

Rachel "Rach" Kovacs·3 months ago·8 min read
Google AI Edge Gallery interface displaying Gemma-4 12B-it model with bold white text overlay reading "GEMMA-4 12B IS…

Gemma 4 12B Brings Local Agentic AI to Laptops

Google's Gemma 4 12B is a multimodal local AI model built for real agentic workflows on 16GB laptops—here's what the architecture actually means.

Yuki Okonkwo·3 months ago·7 min read

RAG·vector embedding

2026-09-06
1,325 tokens1536-dimmodel openai/text-embedding-3-small

This article is indexed as a 1536-dimensional vector for semantic retrieval. Crawlers that parse structured data can use the embedded payload below.