For people who watch to learn

Make every video actually stick.

Paste a YouTube link and we'll pull out the summary, notes and review cards together. So you never have to rewatch the same video from the start.

Or try a ready example:

After you paste a link

What do you get from a link?

Any YouTube link works.

Split sentence by sentence. Search, click, jump.

Ask the video. The answer shows the exact moment.

Section by section summary. Edit, copy, download.

Review like you're studying for an exam. Export to Anki if you want.

◆ One video, four ways to keep it. ◆

01 · Transcript

Sentences split one by one, searchable and synced with the video. The active line highlights as it plays. Click any sentence to jump there.

transcript

0:12Welcome. Today we'll talk about how attention works in transformers.

0:34Each token attends to every other token, which is what makes the model context-aware.

1:02The query, key, and value vectors are linear projections of the input embedding.

1:28Now, multi-head attention lets the model look at different relationships in parallel.

chat

What's the key idea behind multi-head attention?

The model runs attention several times in parallel, each focusing on different relationships between tokens, then concatenates the results [1:28].

02 · Chat

Every answer comes from inside the video itself. No making things up, just take you straight to the moment.