AI study notes from your own documents
A learning claim should show which of your documents it came from. That is the whole idea behind source grounding, and it is the difference between notes you can revise from and notes you have to double-check line by line.
What "source-grounded" means here
It means the Notes are generated from one specific document in your project, and nothing else. Not from the model's memory of similar material, not from the web, not from another project of yours. If a fact is not in the source you uploaded, it does not belong in the Notes.
This is a narrower promise than most AI study tools make, and narrower on purpose. A tool that blends your chapter with everything it has ever read produces smoother notes and gives you no way to tell which sentences are yours.
How a claim points back at your document
When Notes are generated, the prompt names the anchors the model may cite, by range, and supported claims carry one. That is what lets a section of your Notes resolve back to the part of the original it came from.
Every project keeps the source itself readable alongside the Notes. When a claim matters — the one you are about to build an exam answer on — you open the source evidence and read the original wording yourself. The Notes are a way into your material, not a replacement for it.
What happens when the evidence is weak
Notes carry a support status. When the evidence in your source holds up, the status reads supported. When it does not, it reads needs review, and the Notes say so on the page instead of presenting a guess in the same confident voice as everything else.
The same rule governs the Quiz. If your source cannot reliably support the number of questions you asked for, you are told the largest size it can support. Nothing is invented to fill a gap.
A flag you can see is worth more than a fluent paragraph you cannot check. The failure mode this design is built to avoid is the one where you revise confidently from a sentence your material never supported.
The limits this design accepts
- English TXT and MD sources, up to 8,000 characters each — the length that can still be anchored reliably
- Material you own or are authorized to use, confirmed by you at upload
- Notes are generated per ready source, not merged across a whole library
- A source that cannot be read is reported as needing review rather than processed anyway
Your material stays yours
Sources, Notes, quizzes, and learning history are private to your account. There is no public sharing surface, nothing is indexed by search engines, and no part of your workspace is shown to another user.
Deleting a project deletes its learning history with it. You can do that yourself, at any time, from your workspace.
Start with one document you already have and read the Notes it produces.