Not train on your work
Your notes are never used to train any model, ours or anyone else's. They are not sold and not shared with advertisers. See the privacy policy.
Most study tools give you somebody else's summary of your subject. als starts from the material you were actually taught, and never leaves it.
Ask a general AI about your unit and it will answer. It will sound certain. It has never seen your syllabus, your lecturer's emphasis, or the three topics that actually came up last year, so a plausible answer and a right one are indistinguishable to you until the paper comes back.
The fix is not a better model. It is a smaller, known set of pages to answer from, and a citation you can check in one tap.
01
Your question is matched against the material filed under that unit. Nothing else is consulted, so an answer can only be built from pages you supplied.
02
Only the relevant extracts go to the model. The answer is assembled from those, in plain language, at the length the question deserves.
03
Every answer names the lecture and page it used. Treat that as the point rather than a footnote: if the page does not say what the answer claims, you have caught it in one tap.
04
If your library has nothing on the topic, als tells you which lectures you do have on file instead of inventing an answer. That refusal is a feature and it took real work.
Your notes are never used to train any model, ours or anyone else's. They are not sold and not shared with advertisers. See the privacy policy.
No citation, no answer. An unsourced claim is exactly the thing als exists to avoid, so it is not allowed to produce one.
Nothing auto-renews. When 30 days are up you drop to the free limits until you decide to pay again. No card sits on file.
Reading, exporting and deleting what you uploaded works on every plan, including after one lapses. Limits pause features, never your own material.