Codex Desktop on macOS flickers generated text while a response is in progress

Summary

Codex Desktop on macOS constantly flickers the generated assistant text while a response is still in progress. The text appears to visually refresh/re-render during streaming, which makes the app uncomfortable and annoying to use. The issue is visible during active generation and is not just normal model latency.

Environment

• App: Codex Desktop for macOS
• Codex app version: 26.513.20950
• App build: 2816
• Codex CLI version: 0.131.0-alpha.9
• macOS: 26.4.1 (25E253)
• Architecture: arm64
• Hardware: MacBook Air Mac17,3, Apple M5, 10-core CPU, 24 GB RAM
• Sensitive identifiers such as serial number, hardware UUID, and local username omitted.

What is happening

• While Codex is generating an answer, the generated text flickers repeatedly.
• This happens during the in-progress streaming state.
• The behavior is visually distracting and makes longer runs harder to read.
• I saw another GitHub report/post mentioning similar flickering behavior, so this may not be isolated to my machine.

Expected behavior

The generated text should stream/update smoothly without repeatedly flickering or visually re-rendering the whole text area.

Actual behavior

The generated text area visibly flickers while the response is still in progress.

Reproduction steps

1. Open Codex Desktop on macOS.
2. Start or continue a local workspace conversation.
3. Send a prompt that produces a multi-paragraph response or a longer in-progress generation.
4. Watch the generated assistant text while it is streaming.
5. Observe repeated flickering/re-rendering of the generated text while the run is in progress.

Notes

• This seems like a desktop UI rendering/streaming issue rather than a model-quality issue.
• I can provide a screen recording if maintainers need one.
• I have not yet isolated whether this depends on long sessions, generated assets, specific models, or a particular workspace.