GPT-5.6 Terra in Codex repeats the same project-specific status messages

Codex
1

Hi OpenAI team,

I’m seeing what looks like a repeated-output / loop issue when using GPT-5.6 Terra in Codex.

IMG_7179
IMG_71791179×864 126 KB

During a normal Codex task, the progress/status messages kept repeating the same two project-specific status lines from my codebase/context. These were related to internal database/entity work in my project.

The same pair of status messages appeared several times in a row instead of showing normal progress.

Environment

  • Product: Codex
  • Model: GPT-5.6 Terra
  • Reasoning effort: Max
  • Client: Codex
  • Approx. time: 2026-07-10 around 7:50 JST

I’m not sure whether the underlying task was actually looping, or whether only the visible status messages were repeated. But the UI output looked abnormal.

Thanks!

2

Sounds like great first feedback about the model.

In past models, especially with Pro, I’ve seen similar churn with repetitive summaries. It’s as if the observer AI of internal reasoning that writes summaries gets triggered on a schedule, while there is not a substantial new milestone to report about.