Research Mode: Binata researches and writes a complete Torah work for you
Today we are launching Research Mode, the largest capability we have added to Binata since launch.
Until now, Binata answered questions: a good, sourced answer within a minute. But some things a chat answer cannot deliver. A properly built yeshiva-style chabura, with an opening difficulty, a core analysis, proofs, practical ramifications and a concluding novellum. An article for a Shabbat pamphlet. A serious introduction to a tractate. Works like these require real research: reading dozens of sources in their original language, comparing positions, building an argument, and only then writing.
That is exactly what Research Mode does. You describe a topic and agree on a plan with Binata, and a team of researchers goes to work in the background. Twenty to thirty minutes later, a complete work is waiting for you: thousands of words, structured along the outline you approved, with dozens of quotes in the original language and precise source references, each of which passed automatic verification against the original text.
We say this carefully but with confidence: to the best of our knowledge, the output of Research Mode goes beyond what any other AI tool can produce today for Torah writing. Not a polished rephrasing of general knowledge, but a work grounded in genuine source retrieval, in the sources' original language, with a real analytical structure. Three complete, unedited examples are linked below. Judge for yourself.
What you actually get
The product of a research run is a complete article that appears in the chat, like a regular Binata answer, only of a different order of magnitude. The three examples in this article are real runs, exactly as the system produced them, with no human editing. Each title opens the full work as a print-ready PDF, exactly as produced by the product's "Printable version" button (the articles themselves are in Hebrew):
- The Nature of the Sin of the Golden Calf in the Rishonim, a conceptual essay: ~5,300 words, 46 quotes and source references, 8 researchers, ~25 minutes.
- On the Nature of the Prohibition of an Unnecessary Blessing, a yeshiva-style analytical article written to the requirements of a Torah writing competition: ~3,800 words, 34 quotes, 12 researchers across three waves, ~26 minutes.
- Be'er HaIshut: Analytical Studies in Rambam, Hilchot Ishut Chapter 1, a systematic, halacha-by-halacha work: ~7,300 words, 59 quotes, 12 researchers, ~33 minutes.
What happens behind the scenes
What makes Research Mode different is that it is not one long AI request. It is a team:
- A research plan. After the clarification dialogue, an outline is set: chapter headings for the work, and a list of sub-topics to investigate.
- A first wave of researchers. Each sub-topic is handed to a separate researcher working in parallel. Each researcher searches the source corpus, retrieves texts in their original language, follows cross-references, and files a findings report. In the runs above, each team performed between 172 and 353 source lookups and searches, processing millions of words of Torah text.
- Additional waves as needed. At the end of each wave, the lead researcher reviews the findings and identifies gaps: a missing quote, an uncovered position, a topic needing depth. It then dispatches a focused follow-up wave. In the Golden Calf research, for example, the second wave was sent to retrieve the full original language of the Kuzari, after the lead researcher judged the summary insufficient.
- Writing. Only after the findings are in is the work itself written: one continuous argument along the approved outline, with every source placed where it belongs.
- Quote verification. At the end of the process, every quote in the article is automatically checked, word by word, against the original text in the corpus.
המחקר נמשך מספר דקות ורץ ברקע — אפשר לצאת מהעמוד ולחזור, המאמר יחכה כאן.
נושאי הלימוד (5/7 הושלמו):
Quote verification: protection against invented sources
Anyone who has tried using AI for Torah study knows the hardest problem: invented citations. A source that sounds perfect, and does not exist.
That is why every research run ends with a separate verification stage, which is not AI at all but a strict mechanical check: every quote in the article is compared word-by-word to the original text. Each source in the article carries a badge:
- Verified: the quote was found in the source, word for word.
- Close to source: found with minor differences (e.g., spelling variants).
- Source unavailable: the text is not in our corpus and could not be checked.
- Not found: the quote was not found in the source. We flag this honestly instead of hiding it.
In the "unnecessary blessing" research, for example, 33 quotes were checked: 29 verified word-for-word, 3 referenced books outside our corpus, and one was flagged as not found. That badge is shown to the reader, who knows to check it themselves. We consider this transparency a basic requirement for a tool meant to serve Torah learners.
כׇּל הַמְבָרֵךְ בְּרָכָה שֶׁאֵינָהּ צְרִיכָה — עוֹבֵר מִשּׁוּם ״לֹא תִשָּׂא״!ברכות ל״ג א:ל״ה
הציטוט אומת מול המקור בספרייה
כׇּל הַמְבָרֵךְ בְּרָכָה שֶׁאֵינָהּ צְרִיכָה — עוֹבֵר מִשּׁוּם ״לֹא תִשָּׂא״!
How to use it, step by step
- Open the "Research" tab above the message box, next to "Chat" and "Edit". It is available to every registered user.
מחקר תורני
תארו נושא רחב — ובינתא תחקור אותו לעומק ותכתוב עבורכם מאמר מקיף מבוסס מקורות.
נותרו לך 3 מחקרים. שיחת הבירור — חינם.
תארו נושא רחב למחקר מעמיק…
- Describe a broad topic, and, importantly, what matters to you: the kind of output (chabura, pamphlet article, shiur, tractate introduction), the style (yeshiva-analytical, conceptual, practical-halachic), the audience and the scope.
- The clarification dialogue is free. Binata asks questions, suggests directions, and presents a research plan: output style, chapter headings, and the list of sub-topics the researchers will receive. This stage costs nothing, so refine the plan over several rounds without worry. In the competition-article example, the user's remark that the work must be "one built argument, not an anthology of commentators" changed the working instructions of the entire team, and it shows in the result.
המהלך הבנוי של המאמר
- פתיחה בקושיה/סתירה — המאמר ייפתח בקושי מוגדר שיניע את כל המהלך.
- חקירה יסודית אחת כשדרה — האם האיסור הוא דין בהזכרת השם או דין במעשה הברכה — הכול ייתלה בחקירה זו.
- כל מקור מובא כראיה, לא כסקירה — הראשונים והאחרונים הישיבתיים ישולבו בנקודה שבה הם מקדמים את הטיעון.
- סיום בחידוש מגובש — יישוב הקושיה הפותחת מכוח החקירה, והעמדת הגדר המחודש עם השלכותיו.
תתי־הנושאים למחקר
- 1.הסוגיות והקושי הפותח — לשונות הגמרא בברכות ובתמורה, ואיתור הקושי שישמש כפתיחה
- 2.שיטת הרמב״ם — ביסוס הצד שהאיסור מדין ״לא תשא״ בהזכרת השם
- 3.שיטת התוספות והרא״ש — ביסוס הצד שהאיסור מדרבנן ודין במטבע הברכה
- …ועוד ארבעה תתי־נושאים
זה הכיוון? אם כן — אמור "צא לדרך" ונתחיל.
- Approve, and the research launches. The run starts only on your explicit approval, and only then is one research run counted against your balance.
- You can close the window. The research runs in the background on our servers. Leave, shut the laptop, come back half an hour later. The progress card shows which researchers have finished, and the article appears in the chat when done.
- Read, ask, continue. The finished article behaves like any Binata answer: sources are clickable, verification badges are shown, and you can keep talking. Ask for a chapter to be expanded, challenge the conclusion, request an adaptation for a different audience. Follow-up questions are billed as regular chat questions. The "Printable version" button turns the article into a print-ready PDF, like the three examples above.
Pricing
Research is not measured in credits. It has its own allowance:
- Premium: one research run per month, included.
- Premium Plus: three research runs per month, included.
- Single purchase: 15 ₪ per run, available to every registered user (no subscription needed).
- 10-run pack: 120 ₪ (12 ₪ per run). Purchased runs never expire.
Three things worth knowing: the clarification dialogue is always free, and a run is only counted from the moment you approve. A run that fails for technical reasons is returned to your balance automatically. The monthly allowance renews every billing period.
Who is it for
Maggidei shiur preparing a weekly shiur. Chabura writers. Community pamphlet contributors. Learners who want a serious introduction before starting a tractate or chapter. And anyone who wants, even once, to see a complete Torah topic laid out before them, from the sources to the novellum.
Open the Research tab and describe a topic. The first clarification chat is on us. We would love to hear what you create. Write to us.