1. Who writes here
Every article carries a named author with a photograph, a role, a biography and links to their GitHub and LinkedIn profiles. There is no house byline and no pseudonym. The authors are the people who build Heym, and you can find both of them on our About page.
This matters because most of what we publish is about software we wrote ourselves. When an article says how a feature behaves, the person writing it usually has the file open.
2. Sources
We apply four rules to every citation.
- Recency. Statistics, benchmarks and research findings come from the last couple of years. Anything older is cited only when it is the origin of an idea, and is described that way rather than presented as current evidence. In a field where a benchmark can be superseded in a quarter, a figure without a date on it is not evidence.
- Primary over secondary. Where a paper exists, we link the paper, not an article about the paper. We read the abstract or the source before citing it, and we do not quote figures out of a search result summary. This rule exists because we have caught search snippets misreporting a paper's own numbers.
- Reachable. Links are checked before publication. A citation behind a wall that most readers cannot open is a citation that cannot be verified, so we prefer a source that can be.
- Attributed in full. Authors, publication and date appear in the reference list at the foot of the article, not just an unlabelled hyperlink.
3. Claims about our own product
Heym is source-available, which means any claim we make about it can be checked against the code. We hold ourselves to that.
- A statement about how a feature behaves is verified against the current source before it is published, not written from memory of how it was designed.
- Numbers we publish about our own system are recomputed from raw data rather than read off a dashboard, and where both exist we check that the two agree before publishing.
- We publish the parts that do not work. If an article describes a control, a default or a limitation of our own software that we think is wrong, it says so, including when that is uncomfortable.
4. Other companies
We do not publish competitor roundups or head-to-head comparisons against products in our own category. When another company's published work is relevant, we cite it by name and describe it accurately, including when it is better than ours. We do not characterise a competitor's product from marketing copy alone.
There are no affiliate links, sponsored placements or paid reviews anywhere on this blog. Nobody pays to appear here and nobody pays to be left out.
5. AI assistance, disclosed
We build AI automation software, so it would be strange to pretend we do not use it. Articles on this blog are researched and drafted with AI assistance, then edited and approved by the named human author before publication. The author is accountable for every sentence regardless of how the first draft was produced.
What AI is not permitted to do here: invent a statistic, invent a citation, or describe a product behaviour that has not been checked against the source. Those are the failure modes we watch for, and they are the reason for the verification rules in sections 2 and 3.
6. Review before publishing
Before an article goes live it is scored against a published content-quality rubric covering clarity, structure, sourcing, originality and evidence, and separately against an on-page SEO checklist. An article that fails a trust check, such as a claim the page contradicts elsewhere, does not publish until the contradiction is resolved.
Articles are cross-reviewed. There are two of us, and a post written by one is reviewed by the other before it goes live, which is why the reviewer named at the foot of an article is never the person who wrote it. The author remains accountable for the piece; the reviewer is a second pair of eyes on the claims, not a co-author.
7. Corrections and updates
When we find an error, we fix the article rather than leaving it and writing a new one. Substantive corrections, meaning anything that changes a fact, a number or a recommendation, are made to the live page and the article's updated date is changed. Typographical and formatting fixes are made silently.
If you find something wrong, tell us at support@heym.run. A correction from a reader is the cheapest quality process there is, and we would rather be corrected than be quoted wrongly.
8. What we do not publish
- Statistics we cannot trace to a named source with a date.
- Performance figures from tests we have not actually run.
- Claims about a competitor's internals that we have not verified from their own documentation or code.
- Content whose only purpose is to rank, with no reader it was written for.