Codex Desktop renderer CPU spikes when a local web app is running after latest macOS update

Summary

After the latest Codex Desktop update on macOS, running a local web app from a Codex conversation causes Codex UI/renderer processes to consume very high CPU. Closing the local website immediately drops CPU usage back down.

This appears to be a Codex Desktop renderer/performance regression rather than normal load from the local web app itself.

Observed behavior

While the local site was open/running, Activity Monitor showed approximately:

• ChatGPT: ~154.6% CPU
• Codex (Renderer): ~150.3% CPU
• another Codex (Renderer): ~120.4% CPU
• Codex (Service) remained low at ~0.9% CPU

The Codex UI became noticeably sluggish.

As soon as the locally running website was closed, Codex CPU usage dropped substantially.

Reproduction

1. Open Codex Desktop on macOS.
2. Use a Codex conversation to create/run a local website (dev server / local preview).
3. Open the local website.
4. Observe multiple Codex Renderer processes climbing to well over 100% CPU each and the Codex UI becoming very laggy.
5. Close the local website.
6. CPU usage drops immediately.

Expected behavior

Opening or running a local development website should not cause Codex Desktop renderer processes to continuously consume multiple CPU cores or make the UI sluggish.

Actual behavior

The local website acts as a trigger for sustained high CPU in Codex Desktop renderer processes. The issue started after today's Codex update.

Notes

• The local website itself may be using a typical dev server / HMR workflow, but the abnormal CPU is in Codex/ChatGPT renderer processes rather than a comparably high-CPU Node/Vite/Next process.
• The immediate CPU drop after closing the website suggests the browser/preview/local-web integration path is the relevant trigger.
• This may be related to existing Codex Desktop high-CPU / embedded-browser / renderer issues, but the reproduction here is specifically tied to a locally running web app and is highly repeatable.

An Activity Monitor screenshot showing the CPU usage is available if maintainers need it.

No source files or tests are identified. Reproduce the issue on macOS by running a local website from a Codex conversation, then profile the Codex renderer and preview or browser integration while the site is open. Done means the local site no longer causes sustained multi-core renderer CPU usage or a sluggish UI.