Prompting 101

How to talk to the AI

The AI is smart, but it can't read your mind (yet). Writing a good prompt is the difference between getting a perfect study guide and getting garbage.

5 min read·Getting started

1. Drop it

Upload all your messy class notes, slides, and past papers.

2. Boss it around

Tell it exactly what you want to extract or learn.

3. Master it

Get a clean PDF or Markdown file ready for the Editor.

Anatomy of a good prompt

A well-structured prompt has three parts. You don't need all three every time, but including all three reliably produces perfect results.

1
The Target — what to keep
Be specific about subject, question type, or section. The more precise, the less noise.
"Extract all calculus questions from Section B..."
2
The Exclusions — what to skip
Telling the AI what to ignore is just as important as telling it what to keep.
"...ignore title pages, answer keys, and the table of contents."
3
The Format — how to structure the output
Especially useful in Markdown mode. Tell it to use headings, number the questions, or group by topic.
"...output as Markdown with a heading per question and LaTeX for all equations."

Choosing an output mode

Before submitting your prompt, pick the output format that suits your workflow.

PDF modeDefault

The AI reads each page to decide if it matches, then the engine slices those raw pages directly out of your source file and stitches them into a new PDF. Diagrams, fonts, and tables are preserved pixel-perfect.

Best for: past papers, textbook chapters, anything with complex formatting.
Markdown mode

Instead of preserving pages, the AI reads and re-writes the content into a clean .md file. LaTeX math is correctly formatted. Use this when you want editable text for Notion, Obsidian, or the built-in editor.

Best for: extracting specific questions, making notes, importing into the editor.

Good vs bad prompts

The vague prompt (bad)
"give me the important stuff"

"Important" is subjective. The AI has no idea which topic, question type, or difficulty level matters for your exam. Expect unpredictable results.

The precise prompt (good)
"Extract all multiple choice questions about cellular respiration from Unit 3. Ignore the answer keys, title pages, and the references section."

Specifies the question type, the topic, the unit, and explicitly excludes noise. This is the level of clarity to aim for.

Real example prompts

Prompt
"Extract every question and diagram from the Mechanics section of this A-level Physics paper. Skip the answer booklet pages and any formulae sheets. Output as PDF."
Prompt
"Pull out all contract law problem questions. Ignore essay theory questions and reading lists. Use Markdown so I can import it into Obsidian."
Prompt
"Split this 200-page biology textbook into three separate PDFs: 'Cell Biology', 'Genetics', and 'Ecology'. Ignore the index and bibliography."
Prompt
"I want every formula, theorem, and worked example from this lecture slide deck. Ignore the class schedule and learning objective slides. Format with LaTeX math in Markdown."

Advanced tips

Upload multiple files in one job. You can attach 5 years of past papers at once and write a single prompt like "Extract all mechanics questions across all these papers." The engine merges matching pages from every file into one output.
Buckets create multiple PDFs at once. Ask the AI to split content into named groups: "Create three buckets: 'Short Answer', 'Essay Questions', and 'Diagram Questions'." You'll get three separate downloads.
Multiple downloads may get blocked by your browser. If you're using buckets, click "Allow" in your address bar when prompted — browsers often block multiple simultaneous downloads.
Markdown mode re-generates content. Unlike PDF mode which copies raw pages, Markdown mode has the AI rewrite the content. This means it can reformulate questions, clean up OCR errors, and properly format LaTeX — but the wording may differ slightly from the original.