Codex struggles to write proper plan to-do's

Support Bug Reports
,
1

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

When I create a plan with Codex 5.2 or Codex 5.2 high, the model often struggles to create proper to-do’s for a plan. It creates a Todo section in the markdown, but Cursor does not detect them as to-do’s in the plan.

Steps to Reproduce

Instruct GPT-5.2-Codex in plan mode to create a plan.

Expected Behavior

The To-Do’s should be properly formatted in the plan view in the dedicated to-do section and not be part of the markdown part of the plan.

Operating System

MacOS

Version Information

Version: 2.4.21 (Universal)
VSCode Version: 1.105.1
Commit: dc8361355d709f306d5159635a677a571b277bc0
Date: 2026-01-22T16:57:59.675Z
Build Type: Stable
Release Track: Early Access
Electron: 39.2.7
Chromium: 142.0.7444.235
Node.js: 22.21.1
V8: 14.2.231.21-electron.0
OS: Darwin arm64 25.2.0

For AI issues: which model did you use?

GPT-5.2 Codex

Does this stop you from using Cursor

No - Cursor works, but with this issue

1 Like
Plan Mode: No Todo generated
3

Hey, thanks for the report. This looks like a known issue. We had a similar report 5 days ago with Codex Extra High: GPT 5.2 Codex Extra High does not use the native TODOs when planning

It was fixed back then by updating to 2.4, but you’re already on 2.4.21, so it’s either a regression or something model-specific.

We need a bit more info:

  1. Does this happen every time with Codex, or only sometimes?
  2. Can you try another model in Plan mode (Claude 4.5 Sonnet or GPT-5 High without Codex)? Do the to-dos get created correctly there?
  3. If you can, can you share the Request ID? (Top right menu in the chat > Copy Request ID)

This will help us figure out what exactly is going wrong with Codex.

1 Like
4

Hi Dean, thanks for looking into it.

About your questions:

  1. No it doesn’t happen every time, and it happens with both complex and short plans. I don’t really see a pattern or difference between affected plans.
  2. Anthropic models don’t have this issue, I was unable to reproduce it with them (Latest Haiku, Sonnet, Opus). A short test with GPT-5.2 High also didn’t reproduce the issue.
  3. The request ID for one of the chats where this occured is cb910db5-4f1f-48f6-9f02-cca1c3665f8a

image
image1538×574 33.3 KB