PARDESWho is behind this, and how we make sure it's right
A few people in Israel who think the page should be in everyone's hand, and who are more afraid of publishing a wrong explanation than of publishing none. This page is the part of that fear that can be written down.
Who we are
PARDES is built by Ariel and a small group of friends in Israel: engineers, and people who learn. It is not a yeshiva, a synagogue or a religious organization, and it is not affiliated with the Pardes Institute in Jerusalem, whose name we share by coincidence of the Hebrew acronym and whose work we respect. It is an education project that happens to be about Judaism's central texts. There is no congregation, no rabbi on staff yet, and no membership beyond the app.
The name is the four levels of reading a text, pshat, remez, drash, sod: the plain sense, the hint, the search, the secret. We work mostly at the first one.
Which texts these are
Naming the edition is the oldest form of honesty in this literature, so: the Gemara, Rashi and Tosafot are the Vilna edition. The vocalized Aramaic and the English elucidation are the William Davidson Edition; the Hebrew elucidation is Rabbi Adin Steinsaltz's; the Mishnah commentary is Bartenura with Tosafot Yom Tov. All of these reach us through Sefaria, under a Creative Commons non-commercial licence, and the free tier is where they live. The Tanakh is the Leningrad Codex text. Every version and its licence is recorded in the app under Texts and credits.
What is made by machine, and how it is checked
The daily explanations, the narration and the illustrations are produced with AI tools. We say that in the app, on every video, and here. What makes it safe to publish is not the model. It is the gate: when an explanation quotes the Gemara, code searches the source text for those exact words, and if they are not there the explanation is not published. A model can paraphrase; it cannot invent a line of Talmud on our pages. The study partner works the same way: it answers from the page and its commentaries, quotes what it relies on, and when the text does not contain the answer it says so.
A named human reviewer, a rav who reads the daily output and whose name stands here, is the next thing we add. It is not here yet, and this sentence will change on the day it is.
When we get something wrong
We fix it in public. A corrected explanation says it was corrected and what changed. If you find a mistake, tell us from inside the app or by email, and you'll see the fix and your name on it if you want it there. One visibly fixed mistake is worth more to us than a hundred pages nobody questioned.
Why the videos show no faces of the sages
Illustrated stories in the app never depict a named rabbi of the Talmud. Unnamed figures and non-Jewish figures may be shown; Rabbi Akiva may not. That is a decision about respect and about not fixing in your imagination a face that nobody has.
The lab behind the app
Underneath PARDES is a research project on the letter structure of these texts, run under strict rules: every hypothesis fixed before the search, every result checked against a control text that should show nothing, nothing claimed from a single witness. None of that is in the app by default, and it will never be on our channels. The daf is the app; the research is the lab. If you're a scholar and curious, write to us.
Things we decided on purpose
- Quiet on Shabbat and Yom Tov, by your location. Nothing pushes, nothing sells.
- Hebrew dates first, everywhere.
- Ads never next to the text, never inside an answer, and never from categories we wouldn't want beside a daf. The money page has the list.
- Free means free. The free tier is the whole app, not a preview.
- Both languages are written separately, not translated from each other.