Your course notes contain the material, but rereading them is not the same as testing what you know. OGAD (Off Grid AI Desktop) lets you build a local study conversation around those notes. You can ask for an explanation, answer a practice question, and check the source without uploading your course files to a cloud AI service.
Download OGAD for Mac or Windows | Desktop releases
What would you like to do with Off Grid AI?
Have a feature or use case you would like us to support? Tell us what you want to do and which device you use.
Write to support@offgridmobileai.co, join our Slack community, or talk to us on Reddit.
Start with one topic from one course. A focused collection gives the model clearer material and gives you a practical way to check its feedback.
What do you need before studying offline?
Install OGAD and download a local text model that fits your computer. Collect readable notes as PDF, DOCX, TXT, or Markdown files. Projects and document search are free core features on supported Mac and Windows computers.
Finish the first file import and search-model setup while connected. A scanned handout needs a searchable text layer or a checked text version first. Keep the original notes for verification.
How do you make a study project?
Create a project for the course, add a small set of notes, and give the model a clear study role. Start each session inside that project so it can use the uploaded material.
- Open Projects > New project, name it after the course, and press Enter.
- In Knowledge & settings > Knowledge base, select Add files.
- Choose the notes and wait for indexing. Leave the relevant files enabled.
- In System prompt, add a short study instruction, then select Save.
- Open Chats > New chat within the project.
A useful instruction is:
Help me study from the uploaded course notes. Ask one question at a time. Wait for my answer before showing the explanation. Cite the source filename for claims from the notes. Say when the notes do not cover something.
These instructions guide the model; they are not a guarantee that it will obey every turn. Check the source when a correction surprises you.
How do you run a useful first session?
Ask for a question on a named concept rather than the whole course. Answer in your own words before asking for feedback.
Ask me one question about the difference between fixed and variable costs, using the uploaded notes. Do not show the answer yet.
After you answer:
Compare my answer with the notes. Show one thing I got right, one point to correct if needed, and the source for that correction.
The expected result is a practice loop tied to your own material. If the model reveals the answer too early, repeat the instruction or ask for a new question.
Use the feedback to choose your next question
A useful study session has an end condition. Instead of asking the model to quiz you indefinitely, choose one concept and plan a short session: explain it, apply it, then identify a common mistake. Keep the course notes open so you can check any disputed correction.
For fixed and variable costs, you might first answer a definition question. Then ask for a new example and explain your classification. Finally, ask which assumption in that example could change the answer. The model's example is practice material; it is not a statement that your instructor used that case.
After each answer, request feedback in a compact form:
Show the relevant idea from my notes, the part of my answer that matches it, and one specific gap. Do not give a score unless the notes provide an actual marking rule.
Write down the gap in your own words. If you confuse the time period or the definition, ask another question on that point before moving on. If the correction has no source support, check the note rather than memorizing it.
At the end, ask for a short recap of the mistakes discussed in this chat. Compare it with your own notes and keep only the useful points. This creates a repeatable practice routine without claiming that the app tracks mastery, predicts exam results, or replaces the course's marking criteria.
How do you avoid studying a model mistake?
Check the cited note and keep the instructor's terminology when it matters. A local model can still produce an incorrect explanation or a misleading example. Source references show where to look; they do not make the answer automatically true.
Ask the model to label material it adds from general knowledge. Keep unsupported extensions separate from what your course actually teaches.
The project document workflow retrieves selected relevant passages. It does not assess your whole syllabus or guarantee complete exam coverage. Use the syllabus as your own coverage checklist.
What if the study session goes off topic?
| Problem | Action |
|---|---|
| The answer uses another course's terms | Keep separate projects for separate courses. |
| A topic is not in the response | Ask a narrower question and check that the relevant notes are enabled. |
| Feedback conflicts with the notes | Ask for the supporting passage and compare it yourself. |
| The model writes long answers | Request a word limit and one question at a time. |
| Offline replies fail | Confirm the active text model is downloaded and local. |
If you use Pro, turn off Include captured memory in the project's settings for a session focused on course material. Earlier chats in the same project can still supply context.
Can you use it without internet or Pro?
Yes. This study workflow uses free local project search and a local text model after setup. Device sync, web tools, and remote models have separate network behavior. To check the offline path, disconnect internet and ask a new question from the prepared project.
Get OGAD, add one topic's notes, and answer one practice question before reading the model's explanation.
Top comments (0)