How to Study From YouTube With AI (My Daily Setup)

Almost everything new I have learned in the last few years started as a YouTube video. Excel tricks that became client work, Python, Blender, tools for my day job as a business analyst. YouTube is my university. But raw watching is a leaky bucket, and I got tired of knowledge draining out by the weekend, so I built a system around it. This is the whole thing, honestly, including the boring parts.

The daily loop, four moves

Collect, do not watch. Through the day, promising videos go into a queue, not into my eyes, and a paid course goes in the same way when I want it as study notes I can search. Watching on discovery is how an afternoon disappears.
Convert in the gaps. The queue goes through VideoDoc while I do other work, and a whole playlist can go in as one document. Each video becomes a PDF and Markdown file on my machine: timestamped words plus the on-screen content as images. My machine does the watching.
Read first, ten minutes. I skim the captures for the skeleton, then read the transcript fast. Reading is more than twice the speed of listening, and skimming text is instant in a way skimming video never is.
Then talk to it. The document goes to the AI, and this is where studying actually happens.

The prompts that pull their weight

Explain the main idea of this video to me like I
know the basics but missed this topic. Then list
what the speaker assumes I already know.

Make 10 questions from this document, easy to hard.
Quiz me one at a time and grade me honestly.
Cite the timestamp with every correction.

I want to apply this tomorrow. Turn the advice
into a checklist in order, and flag any step the
video skipped over too quickly.

The timestamp citations are the trust anchor. When the AI says I got question 6 wrong, the timestamp takes me to the exact minute of the video to see for myself, so the AI stays a tutor and never becomes an oracle.

The honest costs

Converting is not instant, minutes per video depending on your machine, which is why the queue runs while I work. The AI chat costs whatever your existing plan costs, the documents add nothing because they are built locally, which is exactly why local tools come out the way they do in my comparison of video to text converters. And the method has a failure mode: converting can become its own procrastination. A document you never read is the same as a video you never watched. The system works when reading stays sacred and everything else stays automatic.

Your study engine

Study at reading speed, remember at AI depth.

VideoDoc is the front half of this whole setup: any video or playlist into one AI-ready document, privately on your machine. $19 once, lifetime, 30 day money back.

Try free in your browser Get Pro · $19 lifetime

Quick questions

Which AI should I study with?

Whichever you already pay for or use free. ChatGPT, Claude, Gemini and NotebookLM all handle a good document well. NotebookLM is especially strong when you load many lecture documents as one notebook.

Is this better than just watching on 2x speed?

For learning, yes. Fast watching still cannot be searched, quoted or quizzed. A document gives you reading speed now and random access forever, and the AI layer only works on text it can read.

Does this work for non-study videos, like tutorials for work?

Yes, identically. Work tutorials, tool walkthroughs and conference talks gain even more, because those are the videos you must reference precisely later.

Steal the loop for one week: queue, convert, read, quiz. If your recall of this week's videos is not obviously better, drop it and you have lost nothing but a try.