Company
How we write
Every page on this site is meant to be something you can act on. This is how we choose topics, check claims and code, use AI tools, and fix mistakes.
Updated Oct 25, 2026
The short answer#
Claims about Belvador are checked against the code and the docs; numbers come from runs we recorded.
Claims about other products cite their public pages with the date we read them.
Code examples are run before publishing, with placeholders instead of real keys.
AI tools help with drafts, translation and some illustrations; a named editor checks every page and is responsible for it.
Mistakes are corrected on the page, with the date of the update.
Choosing topics#
We write about problems people bring to error tracking: a cron job that stopped, an error that only happens in production, an AI agent that fails quietly, a bill that grew with traffic. A topic needs a real question behind it and something we can show — a command, a configuration, a measurement.
Checking claims#
Belvador's behaviour is checked against the source and the docs. Limits and prices are rendered from the same configuration the product enforces, so they cannot drift from it.
Other products are described from their own documentation and pricing pages, with links and the date we read them. Comparisons say where the other product is the better choice.
Numbers come from runs we recorded, with the machine, version and date (benchmarks). If we have not measured something, we say so.
Testing code#
Code is run against a test project before it is published, and the page notes the versions it was run with when they matter. Keys, tokens and email addresses in examples are placeholders such as <public-key>. Setup steps on product pages are generated from the same source as the docs and the app, so a step shown here is the step the app shows.
AI tools#
We use AI tools to draft, to translate between English and Vietnamese, and for some illustrations. Every page is then edited and checked by a named editor, who is responsible for it. We do not write to pass AI detectors; we edit for accuracy and clarity. The Belvador logo in images is always the real logo file, never drawn by a model.
English and Vietnamese#
The two versions are written for their readers, not translated word for word. Terms developers use in English — DSN, issue, stack trace — stay in English in Vietnamese pages.
Links#
Links go to the docs page that explains a step, the product page that describes a feature, and primary sources: specifications, vendor documentation, release notes. We do not buy links or place paid links in articles.
Corrections#
If something is wrong or out of date, write to hello@belvador.dev with the page address. We fix it on the page and change its update date. Security issues go to the address on the security page.
Posts on this topic
All postsTry it on your own Hetzner project
Free during the beta. Hetzner bills the server; Belvador never bills per event.