We would rather tell you this plainly than have you wonder.
The short version
The words of the scripture and the literal translations beside them are never written by AI. The plain-English layer and the deeper commentary are drafted with the help of AI, under a written interpretive framework, and then read, corrected and refined by people. Every page says who translated its literal line, and every Hindu page says whether a reviewer has signed it off yet.
What AI never touches
The original script. Gurmukhi and Devanagari are copied exactly from established sources and checked character by character. Nothing in the original script is ever typed from memory or generated.
The literal translation of the Guru Granth Sahib. That is Dr. Sant Singh Khalsa's English translation, reproduced as published.
The transliteration of the Guru Granth Sahib, reproduced as published.
What AI helps with
In plain words — the verse rendered in simple modern English, in the verse's own voice, so a ten-year-old can follow it.
The deeper layer — context, word meanings and connections for the adult reading alongside.
The takeaway and the question to sit with.
These drafts are made with Anthropic's Claude models, working from a written framework that sets the vocabulary, the voice and the rules — for example, that the plain-words layer must carry every line of the verse and must never talk about the verse instead of saying it.
Who checks it
Sikh texts. A person reads the drafts, corrects them, and refines them through real conversations with children and families. The work is version-controlled, so every change is recorded.
Hindu texts. These are newer. Every line of Devanagari has been checked against two independent printed sources. No public-domain English translation exists for most of these hymns, so the literal translation is our own draft, and it says so on every page. Each Hindu verse shows "Draft — awaiting reviewer" until a reviewer has read it; that line changes only when they have.
When we get it wrong
We will. Every verse has a Suggest a change button. Suggestions are read by a person, and corrections are made in the open.
What we will not do
No ads. No paywall. Nothing is generated while you read — every page was written and checked before it was published.
We would rather tell you this plainly than have you wonder.
The short version
The words of the scripture and the literal translations beside them are never written by AI. The plain-English layer and the deeper commentary are drafted with the help of AI, under a written interpretive framework, and then read, corrected and refined by people. Every page says who translated its literal line, and every Hindu page says whether a reviewer has signed it off yet.
What AI never touches
What AI helps with
These drafts are made with Anthropic's Claude models, working from a written framework that sets the vocabulary, the voice and the rules — for example, that the plain-words layer must carry every line of the verse and must never talk about the verse instead of saying it.
Who checks it
When we get it wrong
We will. Every verse has a Suggest a change button. Suggestions are read by a person, and corrections are made in the open.
What we will not do
No ads. No paywall. Nothing is generated while you read — every page was written and checked before it was published.