Codex Desktop: Persian RTL Markdown table columns render misaligned

What version of the Codex App are you using (From “About Codex” dialog)?

Unknown (not captured in the screenshot)

What subscription do you have?

ChatGPT Plus

What platform is your computer?

Windows desktop app

What issue are you seeing?

Markdown tables containing Persian (RTL) text render with visually misaligned columns in the Codex desktop app.

In the affected table, the header labels appear in one visual order, but the body values do not consistently sit beneath their corresponding headers. In particular, numeric values such as word counts (e.g. 2,100, 2,150, 1,200) appear under the wrong heading instead of "تعداد کلمه متن" ("text word count"). The result is difficult to scan and can lead to values being attributed to the wrong column.

The table mixes RTL Persian text, LTR domain strings (for example .com, .net, .org), and numbers. It looks as if directionality or column layout is being applied inconsistently across the header and rows.

What steps can reproduce the bug?

1. Open Codex Desktop on Windows.
2. Ask Codex to render a Markdown table whose headers and prose values are Persian / RTL.
3. Include LTR fragments in cells, such as English domain suffixes, URLs, or inline Latin words.
4. Include numeric columns such as scores and word counts.
5. View the rendered table in the Codex conversation.

Example:

| سایت | موضوع/کلیدواژه اصلی | کلمات کلیدی فرعی شاخص | تعداد کلمه متن | امتیاز مخاطب |
| --- | --- | --- | ---: | ---: |
| میزبان‌فا | انواع دامنه | پسوند سایت، دامنه‌های .com .net .org | ۲,۱۰۰ | 7.5 |
| پارس‌پک | پسوند دامنه چیست | TLD، gTLD، ccTLD | ۲,۱۵۰ | 7 |

What is the expected behavior?

• Each rendered body cell should line up directly under its corresponding header.
• The numeric word-count values should render under "تعداد کلمه متن", and the scores under "امتیاز مخاطب".
• RTL Persian content should coexist correctly with LTR strings and numbers.
• Column order should be stable and consistent between the table header and every row.

Additional information

This appears to be a specific RTL/BiDi table-layout problem rather than simply text alignment. The table is structurally readable as Markdown, but the visual presentation makes the column-to-value mapping ambiguous.

A screenshot of the issue is available from the reporter and shows several rows where the numeric values visibly do not align with the expected header columns.

Related RTL reports:

• #19504
• #26250
• #31903

Start at the Codex Desktop Windows entry point and reproduce the supplied Persian Markdown table, including mixed RTL/LTR text and numbers. Compare the rendering with related reports #19504, #26250, and #31903. Done means headers and every body row preserve stable column mapping, with word counts and scores under their intended headings.