Method
How we publish
Every scripture, hymn and guide on this site is held to the same few rules. They are stated here once so a reader can check any page against them. The work is done by Gaurav Verma.
1. The received text is reproduced verbatim
A verse is printed as the tradition hands it down. We never emend a mantra, never "correct" a reading we find odd, and never silently harmonise a quotation with the text it quotes. Where a reading is doubtful it is kept exactly as received; the doubt is worked out in the apparatus, not in the verse.
Each unit is collated against two independent printed witnesses before it is called done. Where the witnesses differ, the reading that is an outright slip in one and a plain form in the other is adopted; where both are possible, the base witness stands and the variant is recorded. A compendium's own recension of a quoted verse is kept even when the source tantra's received text differs — it is quoting a different recension, not misprinting.
2. Every śloka carries a word-by-word gloss
The padapātha splits the verse into its words in chant order, with a short meaning in English and in Hindi for each. It is a derived layer, kept outside the language switch: the dictionary of a verse's words is never hidden. Sandhi is resolved and compounds hyphenated in the gloss — that is analysis of the verse, not a change to it. A verse whose word-split is genuinely unsettled carries no padapātha rather than a guessed one.
3. The translations are ours, in plain language
English and Hindi are written for a reader with no Sanskrit, in short everyday sentences. Literal fidelity lives in the padapātha; the translation's job is to be understood. The two languages are written to the same standard — Hindi is not an afterthought to the English.
4. Every claim is graded against a verse
When a guide, a note or a post says "the text says X", that sentence is attached to the exact passage and given one of four grades. A grade that says the practice is not in the text is published with the same weight as one that says it is — those rows are the reason the mechanism exists.
- Stated 374
- The passage says this in so many words.
- Inferred 13
- The passage supports it, but a reader has to put two things together.
- Customary 23
- A widely kept practice that the text checked does NOT prescribe. The note says what was looked at and not found.
- Contradicted 37
- The text checked says the opposite of what is commonly claimed.
A cited verse shows a small count on its chapter page; the claim itself lives on the page that makes it, so the received text is never buried under commentary.
5. Living teachers are quoted, attributed and timestamped
Beside the texts, the site indexes what practitioners teach in recorded discourses. Every such note names the source and the second at which the claim is made, so it can be checked against the recording. A teacher's statement and a text's statement are set side by side and graded independently; where they disagree, the disagreement is shown, not smoothed over.
6. Where AI is used — and where it is not
Language models are used as drafting and checking tools: reading scanned pages, proposing a first word-split, drafting a translation for review, comparing two witnesses, and flagging a verse whose metre breaks. Nothing an AI drafts is published without a human reading it against the source.
- No verse on this site is generated by a model. The Sanskrit is transcribed from printed witnesses and rendered from our database.
- There is no chatbot. A page that quotes a text links to the verse; it does not paraphrase it from memory.
- Each chapter records the day it was published and, if its text has been corrected since, the day of the last revision; both are shown in the page's metadata.
7. Corrections
If you find a reading, a gloss or a translation you believe is wrong, write to [email protected] with the chapter and verse. Corrections that check out against the witnesses are made and the chapter's revision date updated.