openai / openai/codex

Codex consumed most of my usage while failing an explicitly approved visual redesign

Open
#43,225 1 comment 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

app bug model-behavior rate-limits
Dominant language
Rust
Stars
125k
Forks
19.5k
PR merge metrics
PR metrics pending

Description

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

5.6 sol high

What subscription do you have?

plus

What platform is your computer?

No response

What issue are you seeing?

my email: panchalrahul521@gmail.com

I need help reviewing a serious Codex task-quality and usage issue.

I asked Codex to transform my existing application into a faithful, end-to-end implementation of an approved “Atlas Operational” visual design. I repeatedly emphasized that every component should match the approved design and repeatedly asked Codex to create the phase prompts so that I would not provide incorrect instructions.

Codex divided the work into seven phases and reported each phase as successfully completed. Approximately three of my five available usage hours were consumed on these phases, including extensive CSS changes, documentation and repeated automated verification.

At the end, the application did not resemble the approved Atlas design. After I challenged the result, Codex acknowledged that it had primarily layered Atlas colours and tokens over the existing legacy interface instead of reconstructing the visual composition to match the approved preview. It also acknowledged that it relied on builds and automated tests as evidence of completion even though those checks did not establish visual fidelity.

This has created significant professional impact. I spent the previous night expecting the approved redesign to be completed, and I committed to my client that the application would be handed to their manual QA team on Monday. I now have very limited weekly Codex usage remaining and an application that is not visually ready for that handoff.

Please review:

  1. Whether any usage assistance, goodwill credit or other accommodation is available.
  2. Why the task was repeatedly reported as visually complete without adequate reference comparison.
  3. Whether the conversation can be reviewed as product-quality feedback.
  4. How I can recover enough Codex access to complete this deadline-sensitive correction.

The repository work is on branch:
codex/client-reference-visual-foundation

The last committed redesign checkpoint is:
565a8b5

Codex’s own final admission was that the result was “a legacy interface with Atlas colours layered on top” rather than the requested Atlas redesign.

I can provide the task share link, screenshots, approved reference image, commit history and QA documents upon request.

What steps can reproduce the bug?

Feedback ID: 01a03d1e-128b-74a3-99fa-aa26e752f3c7

What is the expected behavior?

No response

Additional information

No response

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by reviewing the Codex conversation associated with Feedback ID 01a03d1e-128b-74a3-99fa-aa26e752f3c7, then inspect branch codex/client-reference-visual-foundation at checkpoint 565a8b5. Compare the resulting application with the approved Atlas reference and supplied screenshots; the issue does not define a repository-level fix or a clear completion condition.

Written by the indexing model from the issue text.

Assessment

Tech stack
css
Domain
ai, design, devtools
Issue type
Bug
Difficulty
5/5
Estimated time
Over a week
Activity status
Active
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.