Summary of "So I stopped using Ghostty..."

Context

The creator describes moving off Ghost (aka Ghosty) — a fast, open-source terminal built on libghosty — to a newer terminal called Semox (subtitles show variants like CMOX/CMUX/Semuk; name inconsistencies are likely caption errors). Semox is built on libghosty by the Manoflow team and fundamentally changed how the creator organizes parallel development work.

Why they switched (problem statement)

The creator found traditional terminal multiplexers and window managers no longer matched their workflow:

Longtime use of tmux and GNU screen no longer fit a highly parallel, branching, hierarchical workflow. Tabs/panes/windows/tmux don’t map well to that mental model.
macOS Spaces and tiling window managers (i3, etc.) shrink or steal space from existing views and don’t reflect how the creator wants to organize multiple projects and parallel tasks.
Need: a different hierarchy and UX that supports many simultaneous, switchable contexts without stealing space from other views.

What Semox does well — features and workflow

Semox provides a project-focused, keyboard-first terminal experience designed for parallel work:

Project-centric sidebar: pin, rename, and reorder project sections so the terminal layout mirrors projects and tasks in the IDE/browser.
Nested panes + per-pane tabs: splits can have internal tabs, letting you group related terminals by task rather than mixing everything into global tabs.
Keyboard-first UX: familiar hotkeys retained (e.g., Cmd+D to split); holding Cmd shows sidebar hotkeys, holding Ctrl switches tabs — enabling fast navigation.
Integrations: Cloud Code / Claude (AI assistant) integration; Semox recognizes terminals tied to those tools and surfaces notifications when interaction is required.
Tmux compatibility: integrates with tmux workflows — the speaker still uses tmux sometimes.
Quick dev workflows: demonstrated running dev servers (bun run dev), git workflows, lazygit → open PR links, and how Semox handles link opening.
Outcome: improved mental model and higher parallel throughput across many projects.

Known issues and limitations

Semox is promising but not without rough edges:

Zsh status duplication bug: status line can duplicate and create new empty lines.
Built-in browser is limited: appears to be a webview without user cookies/extensions; links can open inside the app and disrupt existing workflows (there is a setting to open links externally).
Cloud/Claude integrations are intermittent: notifications or tool usage may be unreliable.
Not final: the speaker expects continued iteration and acknowledges Semox introduces new problems while solving others.

Demo and practical tips shown

Practical shortcuts and behaviors demonstrated in the video:

Cmd+D splits a pane 50/50.
Sidebar items can be renamed and pinned to the top for quick access.
Hold Cmd to reveal hotkeys for sidebar items (e.g., Cmd+2, Cmd+3).
Hold Ctrl to switch between tabs within a pane.
Keep dev servers in the same project-pane context (example: bun run dev desktop).
If you prefer external browser behavior, disable “open terminal links in the app.”

Vision and suggested next steps for dev tools

Ideas for advancing the app-level developer experience:

Paper window manager: an app-level “paper window manager” where panes are independent, infinitely nestable and scrollable — opening new panes adds cards you slide between instead of stealing space.
Integrate full Chrome profiles: support extensions, cookies, passwords, and embed non-terminal UIs (VS Code panes, browser pages) into the same high-level workspace.
App hierarchy aligned with human task models: project → task area → pane/tab, with fluid left/right/up/down sliding (paper-like navigation).
Better GUIs for AI agents: current GUIs are immature; improved integrations and open APIs could unlock better agent management.
Community experimentation: encourage folding parts of this vision into tools like T3 Code or other developer tooling projects.

Sponsor

Sponsor mentioned: Milkstraw — an AWS cost optimization service that claims up to 50% savings and charges 20% of the saved amount. The speaker used it to find savings in their AWS bill.

Takeaway / recommendation

Semox is the speaker’s current daily driver and worth trying if you do heavy, parallel, “agentic” development work on macOS.
It’s not perfect, but it introduces a promising hierarchy and app-level approach to organizing parallel development that many existing tools don’t support.
The space is early; there are many opportunities for teams to build on libghosty and experiment with better UIs.

Main speakers / sources mentioned

Video creator / narrator (long-time terminal/tmux user; likely “Theo” from transcript context)
Manoflow team (developers of Semox; speaker disclosed being an investor)
Lawrence (person at Manoflow mentioned)
Mitchell (creator of Ghost/Ghosty / libghosty)
Claude / Cloud Code (Anthropic)
Brody (referenced for Neri/window manager demo)
Igor (comment referenced about macOS Spaces)
Neri (the “paper window manager” concept/project)
Sponsor: Milkstraw (AWS optimization)

Note: multiple names and product spellings in the subtitles are inconsistent (Semox / CMUX / CMOX / Semuk), but the core points above reflect the technology, features, demo, problems, and future direction discussed in the video.