In His Name, the Most High
Questions
Plain answers to what people ask us most.
About IlmFlow
What is IlmFlow?
A set of tools for the students and teachers of the hawza, and a model that learns the hawza’s method from them: how a scholar reasons from the sources to an answer, with its references, and every machine output labelled as one. The tools are in use today; the model is being built.
Does it act as a marja‘ or a mujtahid, giving rulings?
No. It is not being built to give religious verdicts. It is being built to show the road to an answer: what the answer depends on, and the texts it rests on. For any binding ruling it sends you to your own marja‘. Ijtihad is a spiritual and authoritative act, and it cannot be delegated to a machine.
Who is behind it?
IlmFlow is a project of Truth Promoters, built by tpcloud.solutions, and paid for by the people who build it. The scholarship belongs to the hawza: its teachers and students write and review what the model learns from. About us.
The model
What will it not do?
It will not issue fatwas or personal rulings. It will not claim to reproduce any particular marja‘’s reasoning. It will not replace study in the hawza, which is relational and grows over years under a teacher. It will not take political sides. It will not settle a genuine difference between scholars by declaring a winner. And it will not handle irshad, du‘a or the inner life of worship. These are written into the material itself, and a record that breaks one is refused at review.
Why does it learn reasoning rather than answers?
Someone who learns a programming language without the discipline behind it can write something that runs without knowing why it works; someone who learns the discipline picks up the next language in a week. The books of the hawza are the same: read with the method, the next book becomes readable too. So what we collect is worked reasoning (a question, its answer and the sources it rests on), not conclusions.
What happens where scholars differ?
It says so. Where qualified scholars differ, it sets out the positions and the sources each rests on, and says which framework it is reasoning within. A model trained with every difference removed would learn something worse than ignorance: confident reasoning with no sense that people disagree.
Can I use it now?
Not yet. The first reviewed training pairs come first, then a first trained version, tested by the hawza’s own teachers. We would rather say nothing about a date than miss one.
Contributing, and who owns what
What is a record?
One question, one answer, and the sources the answer rests on, with the person who wrote it and the person who checked it both named. It is a worked teaching example of the kind a teacher gives a student who asks, not a passage copied from a book. It is written in Daftar.
Who checks the records?
A second person, always: a teacher, or a student at the level of dars-e kharij, who asks whether the reasoning is sound and whether the source really says what it is claimed to say. A record can be approved, held while a page is checked, sent back, or refused, and always with a reason its author can act on.
Who owns my recordings and records?
You do. Your recordings, their transcripts and the questions and answers you write stay yours: we are their custodian, you can download them for life, and if Daftar ever winds down, you get them before it does. What the project owns is the assembled, reviewed dataset as a whole. It is not a public download, but it is not closed to the hawza: if an institution of the hawza would like it, that is a conversation we want to have.
What do contributors get?
Their own lessons back, recorded and kept for them; credit for their work, under the name they choose; and access for as long as Daftar exists, even if they stop contributing. Those who contribute substantially will have free lifetime access to IlmFlow at a founding tier when it is ready; we state the threshold plainly before anyone starts.
How do I take part?
Every tool is by invitation or on request. Write to us and say how you would like to help: as a teacher, a student, a reviewer, or a publisher whose books belong in Fihrist.
Privacy and security
Does this site track me?
No. It makes no request to anyone else: no analytics, no tracking, no fonts or scripts from elsewhere. It sets one cookie, and only if you choose a language. A message you send us is emailed to us and not stored.
Is there a password to sign in to the tools?
No, there are no passwords. You sign in with a passkey (the fingerprint, face or PIN you already use to unlock your phone or computer) or with a one-time link sent to your email, which stays available to everyone, always. How passkeys work (opens in a new tab).
Can you see my fingerprint or my face?
Never. They do not leave your device and never reach us: your device checks that it is you and tells us only that it is satisfied. What we keep can confirm your device’s answer but cannot be used to sign in as you, so even stolen records would let nobody into your account.
Is my recording sent to an AI company?
No. Our model runs on our own machines, and nothing a contributor records or writes is sent to an outside AI service. The transcription of lessons, when it comes, will run there too.
Can I delete what I have given?
Yes. You can delete your own recordings at any time, and a deletion is permanent. For anything else you would like removed, ask us. Our backups hold what was deleted until they expire, at most about six months later.
Where are the details?
Daftar’s privacy notice (opens in a new tab), terms (opens in a new tab) and roles (opens in a new tab) set them out in full, in English, Farsi and Arabic.
Still have a question?
Ask us anything we have not answered here. The hawza’s own words are explained in the glossary.