// the take
OmniWM (renamed from Hiro) is a macOS tiling window manager written in Swift that implements both Niri-style scrolling columns and Hyprland-style dwindle/BSP layouts. It targets developers on macOS 15+ who want Linux-style tiling WM workflows without disabling SIP, and it ships signed/notarized binaries which matters for MDM-managed corporate machines.
- Notarized and developer-signed releases with no SIP requirement is a genuine differentiator from competitors like yabai — this is what actually matters for anyone on a managed enterprise Mac.
- The codebase is well-structured with clear separation: layout engines (Niri/Dwindle), reconcile/planner layer, AX layer, IPC layer, and UI all in distinct directories. The reconcile/snapshot pattern for state management looks thoughtful rather than bolted-on.
- Unusually solid test coverage for a young WM — over 60 test files covering layout engines, IPC routing, hotkey handling, animation, and the reconcile pipeline. Most tiling WM projects have zero tests.
- Built-in Ghostty libghostty integration for the quake terminal is a sharp feature — you get a real terminal embedding rather than a hacked floating window, with proper tab/pane support inside the dropdown.
- macOS 15+ only (Sequoia) is a hard requirement. Anyone still on Ventura or Sonoma for stability reasons — common in enterprise — is completely locked out despite the enterprise-friendly notarization pitch.
- Building from source requires you to compile Ghostty yourself and manually copy a universal libghostty.a into the right xcframework path. This is a real friction point for contributors and anyone who doesn't want the quake terminal but is forced to deal with this dependency anyway.
- Private API usage (PrivateAPIs.swift, SkyLight bindings, CGSEventObserver) is necessary for responsiveness but means any macOS point release can silently break core functionality. There's no documented fallback strategy or compatibility matrix.
- The README badges 'Claude Code - Assisted', which combined with the rapid feature accumulation and AI-assisted test generation raises legitimate questions about whether the test suite actually catches real bugs or is just coverage theater. Worth auditing before trusting it in production workflows.
We read the README, the directory tree and a key source file in June 2026. Nobody pays to be listed. How we judge →