All other languages

The insights on this site were discovered in English — but they are not for English speakers alone. The crises we face are global. The solutions must be global. And every human being on this planet deserves access to these ideas in their own language.

This site currently supports 10 languages in principle: English, German, Spanish, French, Portuguese, Russian, Arabic, Hebrew, Hindi, and Chinese, but more can be added (see below).

The English version is the most carefully maintained original from which all translations derive — not because English is superior, but because it happens to be the language in which LLoL has been working and in which these structures were first articulated.

Every language folder you see on this site — /de/, /fr/, /ar/, /zh/, and all the others — represents a commitment: no one should be excluded from understanding what is at stake, or from contributing to the solution, because of the language they speak. But to make that happen will require the general support and specific input from a great many people.

Where we are now

AI-powered translation has made stunning progress in the last decades and can now produce serviceable translations at surprisingly cheap costs.

However, translating is complicated ­ as known by all who know more than one language well. Much nuance can be lost in translation and the impact can range from minimal to catastrophic, depending on the context and the details. Therefore, it pays off to have more experienced translators work on translating important texts, because they bring a deeper understanding of both the languages and the content matter to the task.

The same is true for AI translations, where the difference in cost between a cheap but serviceable starting translation to the highest quality translation is about 50-fold with various meaningful stages in between.

That initial draft AI translations are serviceable means that they are generally accurate in meaning (although watch for exceptions!), but they are not as polished in style as higher quality AI translations. Also, precision of expression increases while overall error rate decreases as AI translation costs increase. But some nuances are near impossible to get right by AI. Therefore, AI offers useful starting points, but not a finished product.

As resources grow, translations can be upgraded through progressively higher quality AI passes and, most importantly, through human review by native speakers who bring cultural understanding that no AI can fully replicate.

How translation quality grows

This project uses the StabilityCode system (LINK to IRONROD-StayVS) to transparently mark how mature each translation is, all next to each other in the raw translation-files from which the final website is produced. Here is how the general StayC codes of LLoL’s IronRod (LINK) map to the specifics of translation quality here:

Code	Stage	What it means
MM	MockupModel	AI first draft (Claude Haiku, relative cost ~1x). Serviceable, usually gets the meaning across. Occasionally produces brilliant short phrasings, but can also produce incomprehensible head-scratchers. This is LLoL’s down payment on the intention to include everyone. Human contributors may use this when they have something they think is worth sharing but they are not sure about the overall quality. To do this they best prefix this with their nickname to own their assessment (e.g. LLoL_MM_v1 for the first version MockupModel by LLoL).
NN	NimbleNonsense	Reserved for rejected translations (AI or human). Preserved as comments to document past work and reasons for rejection to help avoid having to re-discuss well-intended ideas that don’t work. All useful ideas in NN ought to have been extracted to the next stage, such that NN remains as a long-term archive of pitfalls.
OO	OddlyOperating	Medium-quality AI pass (Claude Sonnet, relative cost ~5x). Natural prose, proper diacritics, target-language quotation conventions. Humanly speaking, what someone would produce who is fluent in that language but not necessarily a professional translator. OO may still contain major gliches in surprising places.
PP	PathProbing	High single-pass AI (Claude Opus high effort, relative cost ~25x). Best prose quality, scripture-aware, worth consulting for human reviewers. This is at the level of what a professional translator may produce.
QQ	QualityQuest	AI translate + review pass (Opus 2-pass [ultra-high effort], relative cost ~50x). Cross-entry consistency checked, scripture quotes verified. Humanly speaking, this is at the level of what one of the world’s best translators would be confident to propose as final draft, albeit before having it reviewed by peers who may have better drafts.
RR	ReviewedRelease	First official release by a human editor. This versioned variant removes the “translation in progress” warning. A competent native human editor selects the best QQ variant. This requirement means that it may take some time to get there, because the only language for which LLoL can reasonably do that is currently German. As much as LLoL would like to accept the help of everybody in this, in order to ensure the consistent high quality of translations without “in-progress” warnings, LLoL will have to rely on recognizable established credentials in order to select trust-worthy editors for this work. Yet, that shall not keep anyone from improving OO, PP, QQ translations, which can serve to build such trust. In absence of a RRv1 release, the highest quality translation below that wins. RRv1 says that there is no pressing need to help translate this page, but by all means, please point out errors and potential improvements - to be reviewed in anticipation of minor patches, feature releases, or even a reorganized version like RRv2.
SS	StableSource	Authoritative, stable translation. The gold standard. Can only be reached after one or more RR releases have been stable for some time and all their features have been integrated. This does not close the book on further improvements but it signals that tons of work have gone into refining this, so unless someone has some groundbreaking new insights, their time is likely better spent improving other pages.

Each version follows the format XXvNrNpN (version-release-patch), so you can always tell exactly how mature a translation is.

How you can help

If you are a native speaker of any of the supported languages and would like to help improve translations, your contribution is invaluable. Even reviewing a single page moves the StabilityCode forward and helps speakers of your language access these ideas more clearly.

For translators and reviewers

The technical details of how the translation system works — including how to run AI translations, how human reviews are preserved, and how to ensure your work is never lost to a subsequent AI pass — are documented in the Translation System Manual.

No special software is needed to review translations. The .po files that hold all translations are plain text files where each entry has the English original (msgid) and the translation (msgstr) side by side. Any text editor works.

If you are interested to help:

To request the adding of a language, please send an email via this FeedbackFlow link (FF-LINK: translate-new-language-request). This may help assess and prioritize translation needs.

To offer human expert help with translating into a given language, please send an email via this customized FeedbackFlow link (FF-LINK: translate-language-offer). While this website was started by LLoL alone, as soon as it catches on and enough people are interested in scaling up ResearchCity, LLoL will be looking to scale up a competent team of translators. Hence, if you can specify your qualifications for translation that may help in the process of scaling up translation teams to reliably serve all languages.

A note on cultural humility

LLoL is acutely aware that translating content about theology, philosophy, and scripture across languages, cultures, and worldviews is one of the most culturally sensitive tasks imaginable. Scripture quotations that resonate in one tradition’s language may need entirely different phrasing in another. Theological terms carry centuries of nuance. Not everything new in theology is necessarily wrong and not everything old is necessarily correct, even though it has become easy to assume that for some people in some areas. Likewise, not everything new is necessarily better, nor is everything old necessarily worse, even though other people have come to assume that. If humanity wishes to survive this presently ongoing Great Filter Tribulation, then we must learn to work together to combine the best of the old and the best of the new for the benefit of all.

This great task requires all kinds of input: the books of the old masters, the latest and greatest research, the questions of beginners, the comments of experts, the practicalities of pragmatists. The manual input of myriads and the help of the machines for getting it done. The AI translations are a bridge — not a destination. They are offered in the spirit of invitation, not imposition. Every culture, every language, every tradition has something essential to contribute to understanding the whole. This site’s architecture is designed so that native speakers always have the last word over any AI output, like Reality must always have the last word over any human opinion. If this site succeeds, it will help the world to become more HuMaNE by establishing new standards for HUman MAchine Negotiation Encouragement, so all can get the best of all worlds. And if you have an idea of how to make this website more HuMaNE, then use this customized FeedbackFlow link (FF-LINK: offer-to-improve-humane-system).