williamjinq
Joined 1 year ago
Joined 1 year ago
I asked Fable to dig into the logs and do a deep research for relevant evidence:
Related reports counted hundreds of Gatekeeper denials for Codex’s own main executable, eventually exhausting syspolicyd file descriptors and causing unrelated notarized apps to be reported as “damaged”:
https://github.com/openai/codex/issues/25882 https://github.com/openai/codex/issues/28071
Quitting Codex and restarting syspolicyd is only a temporary recovery. The useful upstream action is to identify the launch/update/watchdog caller, add exponential backoff plus a hard retry cap, and publish the first known-fixed desktop build.
I use Claude Code, Codex and OpenCode together, but I wanted a desktop workflow that wasn’t tied to one of them. I kept coming back to my terminal and keep forgetting which agent had finished what. Especially, I want to manage different cron jobs together (Claude Code is better at frontend so I want it to write SEO pages, while Codex is better at logic so I want it to review PRs).
So I built Cetus, with support for different runtimes, and many other stuff(quick launcher allowing you to use it using a global shortcut, cron jobs for different runtime, and kanban view). Under the hood it's a Tauri app with a Rust host.
I’d especially like to know: Do you also switch between multiple agent runtimes, or have you mostly settled on one? Does Cetus make sense to you?