Question Generation

Track chunks to generate AI-powered questions. Choose between open-ended, MCQ, and LeetCode-style formats.

How tracking works

Tracking is how you tell Temar, "I want to review this material." When you track a Chunk, Temar's AI reads its content, generates review questions with detailed grading rubrics, and begins scheduling those questions for spaced repetition reviews.

You can track content at any level of the hierarchy:

• Track a single Chunk — click the Track button on any Chunk to generate questions for just that piece of content.
• Track a Note — tracks all Chunks within the Note at once, using each Chunk's individual configuration.
• Track a Topic — tracks every Chunk across all Notes in the Topic. This is the fastest way to set up reviews for an entire subject.

When you track a Note or Topic, Temar processes each Chunk individually — so every Chunk still gets its own tailored questions based on its specific content.

What the AI generates

For each tracked Chunk, the AI reads the full content (including formatted text, code blocks, math, and diagrams) and produces:

• Questions — targeted prompts that test your understanding of the Chunk's material
• Grading rubrics — detailed criteria the AI uses to evaluate your answers later, including key points you should cover and common misconceptions to watch for

This means every question comes with a built-in standard for what a good answer looks like — making the AI's feedback during reviews both accurate and educational.

Question types

Temar supports three question formats, each testing a different dimension of understanding:

Open-ended questions

Free-text questions where you write your answer in your own words. These are the most versatile format and the best test of deep understanding. During review, the AI evaluates your answer semantically against the rubric — it checks whether you covered the key points, identifies strengths and gaps in your response, and provides specific feedback.

Open-ended questions are ideal for conceptual material, definitions, processes, comparisons, and any content where explaining in your own words demonstrates mastery.

Multiple choice (MCQ)

Four-choice questions with one correct answer. The AI generates plausible distractors — wrong answers that are realistic enough to challenge your understanding, not obvious throwaway options. Each question includes an explanation of why the correct answer is right and why the distractors are wrong.

MCQs evaluate instantly on the client side — no AI call needed during review — so they are fast to answer and do not cost any Pass credits at review time. They are great for testing recognition and distinguishing between similar concepts.

LeetCode-style coding questions

For programming content, Temar generates coding challenges complete with a function prototype, test cases with expected inputs and outputs, and constraints that define the problem boundaries. These questions test whether you can apply the concepts in your Chunk to solve concrete problems.

LeetCode-style questions are perfect for data structures, algorithms, language features, design patterns, and any content where writing code is the best proof of understanding.

Configuring question generation

When you click Track on a Chunk, a configuration popover appears where you can customize the generation:

• Question types — select which formats to generate (open-ended, MCQ, LeetCode-style, or any combination)
• Question count — how many questions to generate for this Chunk. The default is based on content length.

As a general guideline, 1 question per 150 words of content works well. A Chunk with 300 words might get 2 questions, while a dense 600-word Chunk might benefit from 4. You can always regenerate with different settings later.

Cascade tracking

When you track a Note or Topic, each Chunk uses its own individual configuration. If you have not configured a Chunk before, it receives sensible defaults based on its content length and type. You can always go back and adjust individual Chunks after a cascade track.

Generation costs

On the managed plan, each question generation deducts a fractional amount from your Pass balance based on actual token usage and the model you selected. The cost varies by content length and model, but most single-Chunk generations cost well under 1 Pass.

Regenerating questions

If your content changes or you want different questions, you can regenerate at any time. When you regenerate:

• New questions and rubrics are created based on the current Chunk content
• Previous questions are soft-retired — they no longer appear in reviews but their history is preserved in your analytics
• The new questions inherit a scheduling baseline from your past performance, so you are not starting from scratch

This means you can freely update your study material without losing your review history or breaking your spaced repetition schedule.

What happens next

Once your Chunks are tracked and questions are generated, they enter the FSRS scheduling system. New questions become due immediately for your first review. Head to Reviews & Spaced Repetition to learn how review sessions work and how the algorithm adapts to your performance.