Source standard
- Primary sources are preferred: official documentation, terms, service pages, standards, public datasets, and regulatory guidance.
- Claims tied to pricing, limits, policies, or product behavior include a checked date when the detail is likely to change.
- A source is placed close to the claim it supports through method notes, tables, or the article source list.
- AI model claims must separate vendor documentation, independent benchmark context, and our own adoption judgment.
Recommendation standard
- Each article must name the best-fit reader and the reader who should slow down or choose something else.
- AI recommendations must explain the task, scoring method, cost boundary, data controls, fallback plan, and what would trigger a model switch.
- Operational responsibilities such as backups, patching, monitoring, cancellation, and recovery are treated as part of the buying decision.
- We avoid absolute labels such as best for everyone, guaranteed uptime, or cheapest forever unless a primary source supports the exact wording.
Corrections and updates
When a material claim changes, the article should be updated with a new date or the claim should be softened. Corrections can be sent to [email protected] with the relevant source link.