The conclusion up front: this site has no independently written English pages at all. Everything is generated as bilingual pairs, and none of it received native review.This section exists so that English readers do not mistake what they are reading for English-original prose.
How it is produced
Every item — topic, measurement, recipe, ledger row — holds paired fields at the data layer: body_zh / body_en, results_zh / results_en. The build picks one side per language. This is not 'write Chinese, then send it for translation'; it is one dataset written in both languages at once.The difference matters: translation tends to leave the two versions unequal in substance, and this does not.
What that actually buys and does not
| Aspect | Achieved | Not achieved |
|---|---|---|
| Substance | Fields pair up one to one; no section is missing on one side | — |
| Numbers | Machine-checkable; a real error was caught (482 should be 486) | — |
| Terminology | Technical nouns stay verbatim (run_code, engines.dsh, client.inject) | — |
| Naturalness | — | No native review; the English reads as technically correct but stylistically translated |
| Semantic accuracy | — | Unverifiable. No native speaker has read it, so whether the phrasing carries the original meaning rests on the author's own judgement |
So what the English pages are good for
- Fine as technical reference: commands, paths, field names, version numbers, and failure signals come from real runs and source, and are language-independent.
- Not a model for English prose: without review, the sentences lean literal.
- When citing, supply the Chinese page too: where a detail is ambiguous, the Chinese page is authoritative because it was written natively.
This is also a concrete instance of not overstating: claiming 'English supported' would imply localisation quality, whereas saying 'translated, unreviewed' is true. If you want a manual to look commands up in, these pages serve. If you want English you can imitate, they do not yet.