Ask

Getting stuck on a term halfway through an explanation is half of what studying alone is like. Going off to look it up somewhere else is what breaks the thread, so you ask right there and get the answer with your lecture in front of you.

About the slide you are looking at

In the viewer, the raised hand in the dock, or the ? key, opens the chat. The voice pauses, the text field takes the player's place and Mina moves up to the header to listen. If you had a word selected in the script, it gets asked on its own: it is the one you did not understand, and having to type it again would throw away half the shortcut.

The answer arrives in writing, bit by bit, with the slide in front of you: an answer that appears as it is written can be read before it is finished. When you close the chat, if the voice was speaking, Mina says "Let's continue" and repeats the paragraph from the start.

The chat belongs to the deck and is saved with it. Changing slides or closing the document does not erase it, and when you come back it is where you left it.

Where each answer comes from

The model gets the slide, what was being explained and your class notes, if they are switched on for explaining. When the answer comes from there, it says so at the top: "From the deck", "From the document" or "From your notes". When the slide is not enough, it answers anyway with what it knows about the subject, without that label. That way you know what was in your material and what was not.

The conversation carries on: "and an example?" is understood in light of the question you just asked. What it remembers is what was asked on the same slide, or your last question if it was a moment ago, not something from twenty slides back about another topic. And if the answer did not land, you can ask it to explain it another way.

About the whole lecture

The Ask card on the document page opens a chat with the whole document, for when the doubt is not about one slide or you cannot remember which one it was on. If the lecture already has a summary, it offers its topics as questions to get you started.

The model cannot read ninety slides at once, so first the relevant ones are found. They are found by words, without the model, comparing each word by its stem so that "filters" finds "filter". Twelve candidates come out, the model picks the ones that actually answer and writes the reply with those, the summary's topics and your notes. Under each answer come the slides it drew on, so you can go and look.

Searching by words works well in a lecture because people ask with the lecture's own terms: "what is Blackboard". What cannot be found that way, like "what is this about", is covered by the summary's topics. And the search is instant: it does not take the machine away from the explanation being generated.

Dictating the question

Both chats, the raised hand one and the one on the Ask card, have a microphone next to the field: you say the question instead of typing it. It is the same dictation as in notes, turned into text on your Mac, and it needs the model that listens to be downloaded. What you said lands in the field without being sent, so you can fix the word it misheard. And if you send with the microphone open, it closes it, turns it into text and sends it.

What you ask is useful later

Asking is the best sign of what is still shaky. The slides you asked about, in either chat, are among the first to come back in Review, and Practice starts with the part of the slide you asked about in the viewer.