Ask a lecture deck what it actually said. Every slide read, slide numbers kept.

A lecture deck compresses badly. Its content is fragments — a heading, three bullets, a chart with no caption — and the connective tissue was spoken. Read linearly it is close to meaningless, which is why searching a deck for a keyword tends to produce six slides that all mention the word and none that explain it.

CarrelAI reads the deck end to end on upload and indexes it by meaning rather than by slide boundary, so a topic developed across four consecutive slides is retrieved as one thing. Ask about a concept and the answer is assembled from wherever in the deck it was actually built up, with the slides it came from named.

A good proportion of lecture material arrives as images — a deck exported to PDF as pictures, a photograph of a slide taken because it was not shared, a scanned handout. On the Study and Pro plans those pages are transcribed automatically during the same read, so a photographed slide is searchable and answerable like any other.

The reason to do this in a workspace rather than a chat box is what you do next. The slide stays open while you redraw its diagram on a canvas beside it, write a sticky note in your own words, or turn "read the chapter this is from" into a task with a due date and the deck's name on it.

Everything you have uploaded can also be asked at once, which is the query that matters at the end of a term: put the term's decks in a folder and ask across all of them, and the answer comes back saying which lecture it was in.