The New Edit Mode: Four Tools That Review Your Torah Writing
Anyone who writes divrei Torah knows: the writing itself is only half the work. Once the argument is on paper, the second half begins — checking that the reasoning holds, verifying that every quote is accurate to the letter, fixing typos that crept in, and organizing the list of sources. This work demands a second pair of eyes — and not everyone has a chavruta free to review every essay before it goes out.
This week we rebuilt Binata's edit mode from the ground up, and at its center are four tools that do exactly that: they go over your document, each from a different angle, and return findings you can fix with a click. In this article we present each tool — how to run it, exactly what you get from it, and what it costs. And most importantly: we show them in action, on a real document.
What edit mode is
Above the message box, next to "Chat" and "Research", sits the Edit tab. Clicking it opens a full writing environment: the document takes up most of the screen, with a side panel holding the tools and the chat. It is a complete editor — headings, lists, tables, full Hebrew RTL support — with Word import and export to Word and PDF. But that is not the point: every word processor has those.
You start in one of three ways: write a new document from scratch, paste existing text, or — the most common path — click "Open in editor" on a Binata chat answer and continue working on it as a document.
מה נעשה במסמך?
ארבעה כלים בודקים את המסמך כולו — ובצ'אט אפשר לבקש כל עריכה בכתיבה חופשית.
כתבו בתיבה למטה מה לעשות במסמך, למשל:
The side panel opens on the tools home: four cards, one per tool. Below them — the chat. In edit mode's chat you ask for anything in free form: "improve the wording", "add sources for the claims", "shorten the opening". Suggested edits are marked in the document with track changes — you see exactly what changed, and accept or reject each one. Nothing enters your document without your approval.
Now — the tools themselves.
The experiment: an essay with five planted mistakes
To show what the tools can really do, we won't settle for descriptions. We wrote a short essay — "On the Nature of the Mitzvah of Rising Before the Elderly": an inquiry into whether "מפני שיבה תקום" is an obligation of action or of the heart, with proofs from Kiddushin, the Rambam, Sefer HaChinukh and the Shulchan Arukh. And into it we planted, deliberately, five mistakes:
- A logical leap — we wrote "and it is necessarily proven" over a proof that can in fact be deflected.
- A typo — a misspelled Hebrew word.
- An inconsistency — the Shulchan Arukh spelled two different ways.
- A corrupted quote — the Rambam's wording with small changes.
- A fabricated quote — a sentence we attributed to Sefer HaChinukh that is not its language.
Then we ran all four tools on the document, one after another. The findings you will see below — the comments, the corrections, the statuses — are the real run's results, verbatim.
Comments & Insights: a lomdish eye on the argument
The first and most interesting tool. Comments & Insights (הערות והארות) goes over what you wrote the way a talmid chacham reviews a friend's writing: it identifies every building block of the argument — sevara, kushya, terutz, reason, distinction, proof, assumption, conclusion — and comments on each.
How to run it: click "Comments & Insights" on the tools home. An opening screen explains what the tool does — and shows the exact price for this document, before anything runs. Only pressing "Start" begins the review.
הערות והארות
עובר על מה שכתבת כדרך שתלמיד חכם מעיין בדברי חברו, ומעיר הערות על הטיעונים — כדי שתוכל לחזק ולשפר את הכתוב
הערה על כל טיעון — כל סברא, קושיא, תירוץ וראיה שבטקסט מקבלים הערה: מבוסס, צריך עיון, או מוקשה
מה שנסתר בין השורות — הנחות סמויות שהטיעון נשען עליהן, וטענות נגד שכדאי ליישב מראש
משפרים ישר בטקסט — לחיצה על הערה קופצת אל המקום המדויק במסמך — עוברים הערה אחר הערה ומתקנים
What you get: within about half a minute, comments start appearing one by one. At the top — an "Overview" of the whole document: the structure of the argument, where its strength lies and where its weakness. Below it — a comment on every claim, each marked with one of three verdicts: sound, questionable (needs strengthening), or flawed (a real difficulty). Clicking a comment jumps to the exact spot in the document, so you move comment by comment and fix as you go.
What happened in the experiment: the tool returned eight comments. It marked our opening inquiry "sound" — "a well-defined chakirah with two clear sides, in the accepted lomdish manner". And it caught the planted logical leap precisely, marking it "flawed":
מבט כללי
מבנה הסוגיא בנוי כהלכה: חקירה בגדר המצוה, נפקא מינא, ראיות מן הגמרא ומן הראשונים, וסיכום להלכה. עיקר החולשה היא בשלב ההכרעה — הכותב קובע "בהכרח" שהמצוה היא חובת הלב, מכח דרשת מעצים עיניו, אך זהו הצירוף בין הראיות שאינו מוכרח.
קושיאהחקירה: חובת הגוף או כבוד שבלב
יש לחקור בגדר מצוה זו: האם הקימה היא חובת הגוף... או שעיקר המצווה היא הכבוד שבלב
חקירה מוגדרת היטב בשני צדדים ברורים, בדרך הלמדנות המקובלת.
ראיהההכרח מדרשת מעצים עיניו
ומוכח בהכרח שעיקר המצוה היא הכבוד שבלב, שהרי תלה הכתוב את הדין ביראת אלהים...
הלשון "מוכח בהכרח" מופלגת מן הראיה: "ויראת" נאמר על המעצים עיניו — כגון דינו של דבר שבלב, ואפשר לדחות שאין זה אלא תנאי במצוה ולא עיקרה. ועוד: הדרשה הקודמת שהביא הכותב עצמו מקישה הידור לקימה שבמעשה, ואיך יאמר עתה שהקימה "סימן בעלמא". ראוי לנסח זאת כראיה לצד אחד הניתנת לדחייה, לא כהכרח.
Note the red comment: the tool did not merely declare the proof weak — it argued why, from the document itself: the previous derashah we ourselves brought compares hiddur to a physical act of rising, so how can we now call rising "a mere sign"? That is a comment a good chavruta would make.
One more important detail: if the tool finds no claims to comment on — no credits are charged.
Proofreading: three checks in parallel
Proofreading scans the document on three planes at once: language errors (spelling, grammar and punctuation in Torah Hebrew), consistency (terms and names appearing in different forms), and style (clarity and phrasing).
How to run it: exactly like Comments & Insights — a card on the tools home, an opening screen with the price, "Start".
What you get: here the output lands directly inside the document. Every suggestion is marked with a wavy underline beneath the words themselves — red for an error, amber for a warning, blue for a style suggestion. Clicking the mark opens a small popover in place: what is written, what is suggested instead, and why — with "Accept" and "Reject" buttons. You can also work through the full list in the side panel, filter by type, and accept or reject all at once.
What happened in the experiment: the tool returned 11 suggestions. It found the planted typo; it noticed the Shulchan Arukh spelled two different ways and suggested unifying; and it added punctuation and style suggestions of its own that we had not planned.
ולצד השני לכארה לא עשה ולא כלום, שהרי אין כאן הידור הניכר לזקן.
לכארה←לכאורה
נראית טעות כתיב במילה "לכאורה" — הצורה התקנית והמקובלת בעברית היא "לכאורה".
עקביותהשלחן ערוך←שולחן ערוך
במקום אחד נכתב "השלחן ערוך" (כתיב חסר) ואילו בשם הפרק נכתב "שולחן ערוך". כדאי לאחד את הכתיב.
Source verification: every quote checked against the book
This may be the most painful problem in Torah writing — and in AI-assisted writing especially: a quote that looks perfect, and is not accurate. Or worse — does not exist.
Source verification goes over every reference and quote in the document and checks each one against the actual text in Binata's library. Every source receives one of five honest statuses: verified (the quote matches the source word for word), nearly exact (minor differences, e.g. plene/defective spelling), quote mismatch (the document's wording differs from the source), could not check automatically (manual check recommended), and source not found. The distinction is deliberate: the tool never "accuses" a quote it simply could not check.
How to run it: the "Source verification" card on the tools home. Here the price appears on the run button itself — "Run check · up to 8 credits" — because part of the cost depends on what is found, and the quoted price is a ceiling.
What happened in the experiment: all six sources in the essay were checked. The verse in Vayikra, both derashot from Kiddushin, the Rambam and the Shulchan Arukh all came back verified — "the quote matches the source word for word". And the quote we fabricated in the name of Sefer HaChinukh? The tool stopped exactly there and flagged it: manual check recommended.
ויקרא יט, לב
הציטוט נמצא במקור כלשונו
ויקרא י"ט:ל"ב
קידושין לב ע"ב
הציטוט נמצא במקור כלשונו
קידושין ל"ב ב
שלחן ערוך יורה דעה רמד, א
הציטוט נמצא במקור כלשונו
שולחן ערוך, יורה דעה רמ״ד
ספר החינוך מצוה רנז
שגיאה בבדיקת המקור — מומלץ לבדוק ידנית
And when the tool finds a quote that does not match its source, it does not leave you with the problem — it has already fetched the source's accurate wording from the library, and offers to replace it in one click:
רמב"ם הלכות תלמוד תורה פרק וציטוט לא תואם
הציטוט במסמך:
"מי שהוא זקן מופלג בזקנה... עומדים בפניו"
נוסח המקור בספרייה:
"מִי שֶׁהוּא זָקֵן מֻפְלָג בְּזִקְנָה אַף עַל פִּי שֶׁאֵינוֹ חָכָם — עוֹמְדִין לְפָנָיו"
The correction goes in through track changes, so it too awaits your approval. You can also select a single passage and verify only it — for 2 credits.
Source index: the list at the back of the book
Anyone who has opened a Torah sefer knows the index at its end: every source cited in the work, organized by section — Tanakh, Chazal, Rishonim, Halakhah. Source index builds such an index from your document, automatically.
The important point: this is not an AI guess. Every reference found in the document is identified against the books themselves in the library, receives its canonical name, and is grouped into the right section. Whatever was not identified is honestly flagged for review, with a note that it is worth checking the reference for errors.
How to run it: the "Source index" card, "Start" — that's it. It is the cheapest tool: 2 credits for a regular document.
What happened in the experiment: the essay's five sources came back grouped into three sections — Tanakh, Talmud and Halakhah — and the Sefer HaChinukh reference (whose quote, you will recall, was fabricated) was flagged for review. One click on "Add at end of document" and the index enters the document, formatted with headings and print-ready:
מפתח מקורות5
הפק מחדש (2 קרדיטים)תנ"ך1
ויקרא י"ט:ל"ב
תלמוד1
קידושין ל"ב ב
הלכה2
משנה תורה, הלכות תלמוד תורה ו׳
שולחן ערוך, יורה דעה רמ״ד
עוד מקורות1
ספר החינוך (מצוה רנז)
מראה מקום אחד לא זוהה בספרייה — ייתכן שהספר אינו בספרייה, וכדאי גם לבדוק שלא נפלה טעות במראה המקום
What all four tools share
Four principles you will find in every tool:
- The price is known up front. Every tool shows the exact cost for your document before running. No surprises — a paid run starts only with your explicit click.
- Results are saved. Closed the document and came back? The latest findings load instantly, free. If the document changed since the check — you will see a clear notice, and decide yourself whether to re-run.
- Every finding leads somewhere. Clicking a comment, suggestion or source jumps to the exact line in the document.
- Nothing changes without approval. The tools suggest; you decide. Every correction goes through track changes or an approval button.
What it costs
All tools are priced in credits — the same credits as chat questions — and the price scales with document length. For a regular document (up to about three pages) these are the prices; longer documents cost gradually more, up to a cap:
- Comments & Insights — 8 credits (long documents: up to 30). No claims found — no charge.
- Proofreading — 6 credits (up to 40).
- Source verification — up to 8 credits (long documents: up to 40). No sources in the document — no charge. Verifying a selected passage — 2 credits.
- Source index — 2 credits (up to 8). No references found — no charge.
- Edit request in chat — 1 credit for a selected passage, 4 credits for a whole-document edit.
And reopening existing results is always free.
Getting started
Open the Edit tab, paste an essay, article or source sheet you are working on — or open a Binata answer in the editor — and click one of the four tools. The opening screen will explain exactly what the tool will do and what it will cost, and only then do you decide.
And if a finding is worth telling us about — a comment that hit the mark, a quote saved at the last moment — we would love to hear: write to us.