How our articles get researched, written, and checked before anyone reads them.
Research and drafting are AI-assisted. We use Claude to scan sources, pull together what happened, and produce a first draft. Every draft then goes into a review queue where a human reads it, checks the claims, and either approves it, edits it, or throws it out. Nothing publishes automatically. There is no scheduled pipeline pushing articles live without someone looking at them first.
We are telling you this because the alternative, quietly generating content and hoping nobody notices, is how you end up with a site nobody trusts. AI-assisted writing with human review is a legitimate way to work. Pretending otherwise is not.
Before anything is approved it has to pass a set of checks. It has to report something we actually did, tested, ran, or read, rather than restating an announcement. It has to name specifics: versions, numbers, dates, benchmarks. Where we tested something hands-on, it has to say what was worse, not only what was better. And it has to be worth a developer's time in a way the README is not.
Drafts that cannot clear those bars get spiked rather than padded. We would rather publish three things a week than seven things that waste your time.
We link to primary sources: the repo, the release notes, the original thread, the paper. Where we quote someone from Reddit, Hacker News, or X, we attribute them.
When we get something wrong, we fix it, note what changed, and update the published date. If you spot an error, tell us and we will correct it.
Display advertising is clearly marked and has no bearing on our coverage. Advertisers do not see articles before publication and cannot request changes.