How do I use AI to help prepare a shiur or class?

AI is genuinely strong at turning a sicha, a maamar, or a sugya into a teachable shiur, as long as you give it the text instead of asking it to fetch the text. Claude has won this comparison in the community's own tests for organizing sources into a class with a teacher's guide and a student handout.

The pattern:

1. 01Bring the actual text. Paste the sicha or article in. Do not ask the model to find it, because that is where it starts inventing.
2. 02Say what you want built. A 60-minute shiur, the audience, the language, and the pieces you need.
3. 03Ask for the artifacts by name. Speech notes, discussion questions, a one-page handout, a printable summary.
4. 04Add the stories. Ask it to include the stories and anecdotes in full rather than summarizing them away.
5. 05Verify every source. Cross-check any citation or mareh makom against the original, because the model will fabricate references if you let it.

A starting prompt:

Here is the text of a talk.
Create structured, comprehensive speech notes with a logical flow.
I want content notes, not a word for word script.
Include all stories and anecdotes in full, and end with five discussion questions.

If you would rather work from audio, transcribe the recording first (Sofer.ai handles a Hebrew and Yiddish mix), then paste the transcript and ask for the same notes.

For searching a large corpus with real citations, NotebookLM is the tool. Upload your sources and ask for many insights at once, because asking for a hundred forces it to pull from far more sources than asking for ten.

More workflows and the tested prompts are on the community wiki: Shiur Prep.