Each reading states what the commentaries hold in common on
that ayah. Where they differ, the reading says nothing. It does not adjudicate, does
not write “some say”, and does not name a school. The purpose is a page a
Muslim of any background can read without being handed a dispute.
The cost of that rule is real and should be weighed: a reader is not
told when a serious disagreement exists. Whether that trade is right is the first
question we are putting to you.
What is mechanically guaranteed
No hand transcription. Every Arabic text, translation and
commentary quotation is pulled from a sealed corpus at build time. A typing error in
scripture is not possible because no scripture was typed.
Verbatim gate. The build refuses to produce a page if any
quotation is not found character-for-character in its source file.
Completeness gate. The build refuses if any ayah lacks a reading.
All 6,236 are present.
Canonical numbering. Kūfan throughout, asserted against the
per-surah table at build time.
What is NOT guaranteed, stated plainly
The readings were written by a language model, and reviewed by the same
model. That is an author proofreading his own book. It catches a great deal.
It is not independent review, which is why you are reading this.
Some readings drop limits their commentaries supply. Known,
measured, and being repaired - see the front page for 4:34.
Sparse coverage. Not every work comments on every ayah. Where a
work is silent the page says so, but a reader may still form an impression of
consensus that is thinner than it looks.
Translations are counted as separate works. Ibn Kathīr in
Arabic, English and Urdu are three entries, because they are three texts that can
differ from one another. This may overstate how many independent voices are present.
What this project does not attempt
It is not a new translation of the Qur’an, and adds no commentary
of its own. It does not rank the commentators, does not decide between them, and does
not issue any ruling. It compresses; it never adds.
The hadith layer, and what a citation is
Beneath each reading sits a folded list of the reports the
commentaries quote where they comment on that āyah. Sunnī and
Shīʿī collections interleave in one list, ordered by how many
editions carry the report, then by recorded grading, then by title. They are
never grouped, coloured or badged by tradition. The collection name tells
anyone who knows; nothing else on the page does.
A citation is a fact about the commentary. It says a
named edition quotes this report in its comment on this āyah. It does
not say the report explains the āyah - that cannot be verified
by this build or by a reader, so it is never written, implied or
templated.
Every link was re-derived before it shipped. A run of
twelve consecutive words has to appear in both the hadith text and that
edition’s comment on that āyah, in the files as they stand today.
Anything that could not be reproduced was dropped rather than shown.
Gradings are quoted, never assigned. Where a named
grader recorded one - al-Albānī, Shuʿayb al-Arnaʾūṭ,
al-Majlisī, Mohseni - it is shown with his name and his own wording.
al-Bukhārī and Muslim carry no per-report grade in this corpus, so
their reports show none and sort as ungraded. That is a fact about the data,
not a judgement about the Ṣaḥīḥayn.
Under an āyah a long report is shortened and the cut is
marked. Its own page carries it whole.
The Urdu has two origins and says which. Where the
publisher shipped an Urdu with the Arabic, that is the source speaking.
Where this project translated one, the report says so in all three
languages.
The repair now under way
Every claim each commentary makes on an ayah is being extracted and
counted. A claim asserted by three or more editions becomes a required element,
and a gate refuses any reading that drops one, naming the editions that assert it. On
ten test ayat this found 53 dropped elements - including, at 4:34, the qualification
that the striking is ghayr mubarriḥ. Your comments will be folded into
the same repair pass.