Weekly window, evidence first

When does the Codex weekly reset window actually start?

People often describe a weekly reset as starting when they log in. That may be an account-specific observation, but sign-in and first quota-consuming use are different timestamps. Use the account's own before-and-after values to test the relationship instead of turning one report into a universal rule.

Last reviewed August 15, 2026 · Independent community reference

1

Capture the account before activity

Record the weekly label, next-reset value, remaining allowance, plan or workspace context, and timezone before the first prompt that may consume usage. Keep short rolling limits and banked credits in separate rows.

2

Separate sign-in from first actual use

A login timestamp and a quota-consuming activity timestamp are not the same observation. Treat the relationship as a hypothesis until the account surface shows which action changed the weekly window.

3

Compare accounts independently

Two accounts can show different weekly dates because their windows may have different anchors, plan contexts, or activity histories. A result for one account should not be promoted as a universal reset rule.

4

Verify after the expected boundary

Take a second sanitized capture after the displayed weekly boundary, then compare the two records locally. If the values do not explain the change, label the result unknown instead of guessing.

Window Anchor Lab · browser-only

Paste three sanitized snapshots to test whether the weekly reset value moved after sign-in or after actual use. The raw text stays in this tab, is never uploaded, and is rejected if token-like fields are detected.

A · Before activity

Before signing in or sending a prompt

B · Signed in, no prompt

Optional: after login, before quota-consuming use

C · After first use

After a small prompt that should consume usage

The three-snapshot method

Snapshot A

Before first use

Save the displayed weekly date, allowance, and limit labels before any activity that could consume quota.

Snapshot B

After a tiny prompt

Record the activity time and capture the same fields again. This separates a visible login from an observable usage change.

Snapshot C

After the boundary

Compare the next-reset value, remaining allowance, and any banked balance after the expected weekly boundary.

Want to compare the records privately?

Paste two sanitized read-only snapshots into the browser-only checker. It does not upload quota data or read your login.

Compare snapshots locally ->

Keep personal timing separate from public resets

A public reset signal can be broad while a weekly, five-hour, or banked window remains account-specific. ResetPulse therefore publishes public source and observation timestamps separately from anonymous account or plan outcomes. The public timeline can tell you what was observed; your own Codex surface remains the source of truth for your personal window.

See a readable propagation trace -> · Open the missing-reset checklist

Use UTC when sharing a result

For a useful community report, include the account or plan context, the exact limit label, the local time and UTC conversion, and whether the value came from before or after activity. If one of those pieces is missing, mark the result as unknown. A public announcement is not proof that every account changed at the same moment.

Check the current public record

Read today's public reset answer, review the usage and banked-reset guide, or inspect the live monitor health. ResetPulse is an independent community monitor and does not require Codex credentials.