Radar/Transparency
Transparency

Codex Reset Radar Methodology

See the sources, evidence tiers, classification rules, and correction policy behind every Codex reset verdict.

Updated July 16, 2026Source-first editorial guide

Source tiers

Primary sources are explicit statements from the Codex lead, OpenAI leadership, official OpenAI accounts, OpenAI Help, or OpenAI Status. Supporting sources include Codex product and developer-experience staff. Community reports are corroboration only.

We watch Tibo Sottiaux first because OpenAI identifies him as the leader of Codex and his account has repeatedly announced quota resets. Sam Altman and official OpenAI accounts are important for company-wide capacity changes and promotions.

Verdict rules

Confirmed requires explicit reset or quota-restoration language and a defined scope. Watching means credible signals exist without a direct promise. No evidence means monitored sources contain no new reset indication. Stale means the data could not be refreshed reliably.

Forecast percentages are editorial estimates, not OpenAI probabilities. A recent hard reset creates a cooldown penalty; direct launch or compensation language increases the score; unrelated outages and vague enthusiasm do not.

Corrections and safety

We link to original public posts, show UTC times, and keep reset types separate. We do not ask users for Codex credentials, local auth files, cookies, tokens, or private usage data.

Material corrections are dated. If a source is deleted or contradicted, the event remains visible with an explanatory note rather than disappearing from history.