Ask once. Answered across everything.
Put a question to the whole collection and TubeOnAI reads every source at once — then shows the references, so you can jump straight to the lecture minute or the page it came from.
Group lectures, papers, podcasts, and articles into one collection. Ask once and get an answer drawn from all of them — with references back to every source.
What are the main types of learning?
Your sources agree on three families: supervised, unsupervised, and reinforcement learning.1
Lecture 4 separates them by the signal the model receives,2 while the podcast argues self-supervised learning is best read as a special case of supervised learning.3
Collections
Group your sources, ask across all of them, and reuse the answers — without opening twenty tabs.
Put a question to the whole collection and TubeOnAI reads every source at once — then shows the references, so you can jump straight to the lecture minute or the page it came from.
Group summaries by topic, project, client, or theme. Build as many collections and categories as you need — there is no cap on either.
Lecture recordings, research papers, podcasts, and articles sit side by side. Add them all to a single collection and read them as one body of work.
Turn a full collection into study notes, a quiz, a blog post, or a newsletter — drawing on every source at once instead of one summary at a time.
Playlist import
Point TubeOnAI at a YouTube playlist and every video is imported as its own entry — the whole course arrives organized, summarized, and ready to question.
How it works
Create a collection, add your sources, and start asking. No setup, no training.
Try it with your sourcesName it by topic, project, client, or theme — however you already keep track.
Drop in lectures, papers, podcasts, and articles — or import an entire playlist at once.
Question everything at once, follow the references, then repurpose what you find.
Your library
Browse a collection and ask across every source from the same screen.
Walks through the update rule and argues learning rate matters more than depth early on.
Compares dropout, weight decay, and early stopping across six benchmark tasks.
Traces the shift from recurrence to attention and what the change cost in practice.
Diagram-first explanation of the chain rule as it applies to layered networks.
Bias-variance tradeoff with worked examples on a small tabular dataset.
Meta-analysis of fourteen studies on compute-optimal training budgets.
Who it's for
However you research, Collections fits the way you already work.
Turn a semester of lectures and papers into one place you can actually question.
Keep a literature review together and trace every claim back to its source.
Build course material from lectures, readings, and talks in a single collection.
Group competitor content by campaign, then repurpose the insights into posts.
One collection per client, so every brief starts from what you already know.
Group your sources into one collection and ask it anything. Every answer comes back with references.
Get started