Codex usage dashboard appears to over-consume weekly quota and may misclassify activity as "Other"

What version of Codex is running?

Codex Web usage dashboard
https://chatgpt.com/codex/settings/usage

What subscription do you have?

Pro

Which model were you using?

Mixed usage through Codex surfaces.
This report is primarily about the usage dashboard / metering, not a single model response.

What platform is your computer?

macOS 10.15.7
Chrome 145

What issue are you seeing?

My Codex weekly usage appears to be draining much faster than expected, and the dashboard is not giving a reliable explanation for why.

Current dashboard state:

• Balance -> 5-hour usage limit: 35% remaining
• Balance -> Weekly usage limit: 24% remaining
• Weekly reset shown: Mar 25, 2026 at 14:56
• GPT-5.3-Codex-Spark 5h: 100% remaining
• GPT-5.3-Codex-Spark weekly: 100% remaining

The problem is that my general weekly bucket is being consumed very aggressively, even though this does not line up with my historical usage pattern.

Additional suspicious details:

• The reset anchor previously behaved differently and has shifted in ways that are hard to reconcile with continuous daily usage.
• The 30-day usage chart shows a meaningful amount of activity categorized as "Other".
• I strongly suspect that at least some activity may be misclassified in the dashboard, because the chart does not make it clear what "Other" actually includes.
• This makes it impossible to verify whether the heavy weekly burn is legitimate task/context consumption or a metering / attribution bug.

In short:
the dashboard seems to be either

1. over-consuming the weekly quota,
2. misattributing usage to the wrong bucket/category,
3. showing a misleading reset anchor,

or some combination of the above.

What steps can reproduce the bug?

I cannot give a guaranteed minimal repro, but this is the pattern I observed:

1. Use Codex regularly on a Pro account across consecutive days.
2. Check Codex -> Settings -> Usage.
3. Observe that the main Balance weekly quota drops much faster than expected.
4. Observe that GPT-5.3-Codex-Spark remains untouched while the general Balance weekly bucket keeps draining.
5. Observe that the usage chart includes "Other" activity that is not clearly explained.
6. Observe that the reset timestamp / anchor is difficult to reconcile with the visible usage pattern.

What is the expected behavior?

• Weekly usage should deplete proportionally to actual metered work.
• The dashboard should clearly explain what counts toward the main Balance weekly bucket.
• The dashboard should clearly explain what "Other" means.
• If CLI / local / app / cloud activity is being grouped into "Other", that should be made explicit.
• The weekly reset timestamp should be intelligible from the visible usage history.
• Users should be able to audit why a large portion of weekly quota was consumed.

Additional information

• I am on ChatGPT Pro.
• I have attached screenshots of:

1. the current usage summary,
2. the 30-day usage chart.

• The issue is not simply "usage varies with task complexity".

I understand that already.
The issue is that the dashboard currently does not provide enough fidelity to explain the observed weekly burn.

• Please verify whether this is:

1. legitimate high consumption,
2. phantom / background consumption,
3. category misclassification,
4. incorrect weekly bucket accounting,
5. or a dashboard/UI bug.

Screenshots

<img width="1104" height="402" alt="Image" src="https://github.com/user-attachments/assets/f434cd75-1deb-4466-abbc-cffd79e8f1aa" />
<img width="1100" height="422" alt="Image" src="https://github.com/user-attachments/assets/ddcf21be-b8fb-41aa-a822-f801e0b319e2" />