We have made a few quiet but important improvements to how Kram AI processes and generates quizzes. This update focuses on two things that matter most: speed and accuracy.
1. Faster uploads and processing
Uploads now complete up to 2x faster for most users, especially with larger PDFs and slide decks.
We rebuilt the parsing layer so Kram AI reads your file in smaller, parallel chunks, which means less waiting after you hit upload.
- Noticeably faster for long documents and class PowerPoints
- Better perceived responsiveness from upload to first question
2. Better note interpretation
Kram AI now recognises structure more intelligently.
- Headings and subheadings are handled separately from main text
- Bullet points are grouped by topic, not line by line
- Lists, definitions, and examples are used as context to improve relevance
- Fewer out of context questions
- Less repetition of similar items
- More accurate targeting of key concepts
3. Cleaner question logic
We fine tuned how Kram AI decides what type of question to generate. Instead of over focusing on recall, it now mixes in application and understanding questions where your notes provide enough detail.
- Better balance across recall, understanding, and application
- Question types reflect the depth available in your notes
4. More consistent phrasing
We updated the internal prompt system so questions are concise and natural. You will see less redundancy and fewer awkward phrases.
- Clear in what it is asking
- Short enough to skim
- Grammatically consistent across a full quiz
5. What is next
We are testing improved support for OCR based files (scanned notes or images) and more context aware explanations for each answer. Both are coming soon.
If you notice faster generation or better question quality, that is this update doing its work.
Kram AI is now quicker, cleaner, and better at understanding your notes. The end goal stays the same: low friction quizzes, high impact learning.