How we operate
Editorial Policy
The rules Codex Insider publishes under: where facts come from, how corrections work, how AI is used, who pays for the site, and what gets labeled sample data. This page is part of the product — hold us to it.
Published
Where do our facts come from?
Primary sources, linked inline. Release facts come from the official Codex changelog and the openai/codex GitHub releases; pricing and limit facts from OpenAI's published pricing pages; error behavior from the issue threads where it was reported. Reference pages carry a visible verification date, and numbers OpenAI reports about itself are labeled as self-reported.
- Every factual claim links to a primary source — a GitHub release, an issue thread, an official changelog entry, a pricing page, or a dated announcement.
- Living reference pages (limits, models,releases, stats) show the date each fact was last verified, so you can judge freshness for yourself.
- Numbers OpenAI reports about itself — weekly users, benchmark scores — are labeled as OpenAI self-reported wherever they appear, because no independent measurement exists for most of them.
- When we can't verify something against a primary source, we leave it out or say plainly that we don't know.
Codex passed 5 million weekly users perOpenAI's June 2, 2026 announcement(verified Jul 9, 2026). That figure traces to OpenAI alone, so wherever it appears on this site it's labeled OpenAI self-reported.
What happens when we get something wrong?
We fix errors in place, with a note in the page changelog. Substantive corrections get a dated entry describing what changed and why; a correction that alters a story's central claim is flagged at the top of the article, not buried at the bottom. We don't silently rewrite published pages.
Spotted something wrong? Reach the desk through the channels on theAbout page and we'll verify against the primary source and fix it — the changelog note will credit the catch.
How do we use AI?
Research and drafts here are AI-assisted, and a human reviews and approves everything before it publishes. We think that's table stakes in 2026, and we'd rather say it plainly than make you guess. AI watches the release feed and writes the first pass; a human decides what's true, what matters, and what ships.
- Nothing publishes without human review and approval.
- AI never invents a number — every figure is checked against its primary source before it appears.
- AI drafting doesn't substitute for reading the source. If the source can't be read, the claim doesn't run.
Who pays for this site?
Nobody — this site sells nothing. Codex Insider has no affiliation with OpenAI, publishes no sponsored content, and carries no affiliate links. There's no product behind the coverage and no vendor buying placement, which means a reference page can call a plan a bad deal without a conflict of interest.
Codex Insider is an independent publication and is not affiliated with, endorsed by, or sponsored by OpenAI. ‘Codex’ is a trademark of OpenAI referenced here for identification purposes only.
What counts as sample data?
Any figure that exists for design or demonstration rather than reporting — a mock headline in a layout, a placeholder chart, an illustrative count — is labeled ‘sample data’ wherever it appears. If a number on this site isn't labeled sample data, it's a reported fact with a source link behind it.