How we write
Editorial policy
Yibud is a small project. This page exists to make the rules of how we write visible — what we check, what we don't publish, and how we handle it when we get something wrong.
Last updated · July 21, 2026
Our promise
What readers should expect
Every article on Yibud is written by the person who built the product. It is reviewed before publication against a short, fixed editorial checklist (see below). Articles are not generated, rewritten, or summarized by an AI language model. They are drafted, edited, and revised the slow way.
The Yibud score reports are generated by a deterministic rule engine — not by a language model. The optional AI layer only polishes the prose; it never decides the numbers, the verdict, or the recommendations. When the AI is unavailable, the structured report is still complete and valid.
How articles are researched
The process behind each essay
Each article is built from a list of specific, named sources — books, founder essays, post-mortems, and primary research where it exists. The reference list is published on the Sources page, and the inputs that drive each essay are listed at the end of the article.
Practical experience is prioritized over theory. When a claim can be supported by what I have shipped, broken, or watched fail, that experience is cited. When a claim comes from a book, the book is cited. When neither applies, the claim is either dropped or marked as opinion.
Examples are verified before publication. Founder quotes, company histories, and revenue figures are checked against the original source. If a number cannot be verified, the article uses an honest qualifier ("by their own report", "publicly stated") or removes the number entirely.
What we don't do
Hard lines we won't cross
No fabricated statistics. If a number can't be verified, the article either doesn't include it or flags the uncertainty explicitly.
No invented founder quotes. Quotes are either verbatim from a public source or not used.
No AI-generated filler. Sentences that exist only to pad length are cut. If a section feels thin, it gets cut, not filled.
No growth-hack theatre. We don't promise "10x your MRR" or "unlock the framework successful founders use." The articles are written for founders who are skeptical of that voice.
No undisclosed sponsorships or affiliate links in the body of any article. If we ever recommend a product, the recommendation stands on its own merits and is flagged at the end of the piece.
Updates & corrections
What happens when we get something wrong
Every article carries a visible "Last updated" line at the top of the body and a full revision history in its frontmatter. Readers can see exactly when an article was published, when it was last reviewed, and which parts have been changed.
Corrections are made in-place — the corrected text replaces the old text, but a note is added to the revision history so the change is visible. We do not silently rewrite articles. The audit trail is part of the publication.
If you spot an error, an out-of-date claim, or an unverifiable citation, please tell us. Corrections are credited on request. Mistakes are a normal part of the work; hiding them is not.
Want to see the engine in action?
Startup MRI applies the same evidence-based reasoning described above to your own idea. Free, no signup, report in under a minute.