Short sessions,
results for years.
Kavke isn’t another generator of text you read once. It turns an AI conversation, a document or a recording into questions, then brings them back exactly when you start to forget. You put in a few minutes a day.
See how it worksNo card. One account — iPhone, Android and browser. Your first course “How we really learn” is waiting after sign-up.
Hi, Anna!
Machine Learning
8 min read, 12 lessons, 48 questions
History of Science
6 min read, 10 lessons, 36 questions
You do the learning. We remove all the friction.
Your part is a few minutes of focus a day — the active recall that builds lasting memory. The material, the questions and the review schedule take care of themselves.
- 01
Add
An AI chat, a YouTube video, a PDF, a photo of a page, a recording or just a topic — Kavke turns it into review questions, with a short lesson underneath.
- 02
Answer
In your own words, by text or voice — in the quiz and at Feynman checkpoints inside the lesson. AI grades the meaning of your answer, not a click on a, b or c.
- 03
Retain
The FSRS algorithm reminds you to review exactly when you start to forget.
Read it once
- AI writes the text, you read it
- A week later a few percent remain
- You start over from scratch
Remember for years
- You answer from memory, in your own words
- The review returns before you forget
- Every session stretches the gap to the next
Start with anything. We’ll handle the rest.
A YouTube video, notes, a PDF, a photo of a page, a lecture recording, just a topic — or a plain chat conversation. Kavke turns it into a ready course with questions. The course isn’t the goal. It’s fuel for questions and reviews.
YouTube
Paste a link — lessons from the transcript, with timestamps
Topic — AI from scratch
Type a topic and AI builds the whole course
Text / notes
Paste an article, script, or your own notes
PDF / DOCX
Upload a document — AI extracts the meaning
Photo (OCR)
Snap a page — OCR reads even handwriting
Voice recorder
Record a lecture or your own take on the topic
Anki import
Upload an .apkg deck — your cards become a course with reviews
In-app assistant
Say what you want to learn — it asks a few questions and builds the course
Say “save this to Kavke” — straight from your chat
Deep in a conversation with Claude, ChatGPT, Perplexity or Mistral? Add the Kavke connector and what you’ve worked out becomes questions in your reviews. The connector only receives what you consciously send.
See how to connectCreate course
Pick a content source. Kavke turns it into a ready course with questions.
Topic — AI from scratch
Type a topic and AI builds the whole course
YouTube
Paste a link — lessons from the transcript, with timestamps
Photo (OCR)
Snap a page — OCR reads even handwriting
PDF / DOCX
Upload a document — AI extracts the meaning
Text / notes
Paste an article, script, or your own notes
Voice recorder
Record a lecture or your own take on the topic
Anki import
Upload an .apkg deck — your cards become a course with reviews
Great explanation! Save this to Kavke.
Done — I created the course “Photosynthesis: light & dark reactions”.
8 flashcards · question types matched to content
In Kavke: course ready
1 lesson · 8 questions · ready to study
We check that you understand — not that you recognize an option.
Every app has flashcards and quizzes. The difference is the quality of the signal entering your reviews: you answer in your own words, and AI grades understanding — not a click on a, b or c.
AI grades the meaning of your answer
- Answer in your own words — AI judges the meaning, not a match against a template
- Ambiguous answer? AI asks a follow-up instead of grading right away
- True/false demands a justification — a 50/50 guess isn’t knowledge yet
System 1 vs System 2
System 1: fast, automatic, intuitive. Recognizes faces, reacts to danger. Works effortlessly.
System 2: slow, conscious, analytical. Solves 17 × 24, controls impulses. Requires focus.
Feynman checkpoint
Explain in your own words: System 1 vs System 2
Heuristics & traps
Anchoring effect
The first number you hear shifts later estimates — even when it's random. In one study a wheel of fortune was spun before a percentage question — a higher number gave a higher answer.
Availability heuristic
We judge how likely an event is by how easily it comes to mind. After a plane crash in the news, we overestimate the risk of flying.
From course: How We Really Learn
How does System 1 differ from System 2 in thinking?
Your answer:
System 1 is fast and automatic, System 2 slow and analytical.
You’ve got the core: fast–automatic vs slow–analytical
Missing: a real-life example
Give an example of a decision driven by System 1.
When I hit the brakes reflexively before thinking.
The gap comes back as a flashcard in reviews
“Explain it” — the Feynman method in every lesson
After each section of a lesson, a short checkpoint: explain what you just read in your own words — by voice or text, as if teaching someone hearing it for the first time. At the end, you retell the whole lesson.
AI scores how complete the explanation is against a rubric — kindly, without nitpicking style.
Why it matters: reading gives you the illusion of understanding. Only saying it out loud, in the simplest words you can, shows what’s really missing — and that effort is exactly what writes knowledge into memory.
Gaps in your explanation come back as flashcards in your reviews.
Questions about what matters
Questions are built from each lesson’s key concepts — never padded to a count
Confidence calibration
Mark how sure you are before answering — and see where confidence drifts from results
No made-up facts
Topic courses are grounded in sources, the AI is barred from inventing facts and dates, and each question shows its source quote
Full features from day one.
You start free, with access to every feature. On your phone and in the browser. No ads, no card.
Free
- 3 courses
- 30 lessons per month
- Phone & browser
- All features
- No ads
Start free
Pro
In-app- Unlimited courses
- Unlimited AI lessons
- Unlimited “Explain it”
- Everything in Free
Buy in the app — a subscription on the App Store or Google Play
Mobile app
- iPhone and Android — App Store and Google Play
- Reviews in a queue, on the bus, on a break
- Listen to lessons (TTS) and answer by voice
Web app
- Runs in any browser — nothing to install
- Big screen — comfortable course creating and editing
- A chat assistant that builds courses
Start on your phone, finish in the browser — same progress, same reviews.
Remember for good, no cramming.
The FSRS algorithm reminds you to review exactly when you’re starting to forget — the moment the memory trace strengthens most. You study less and remember longer.
FSRS algorithm
A newer standard than Anki’s SM-2 — more retention, fewer hours
You know when they return
After a quiz you see when each question comes back — nothing gets lost
Mixed reviews
Questions from different courses, interleaved — a true memory test
Memory classification
Long-, medium- and short-term — you see what’s already solid
Reviews
25Questions from all courses — research shows interleaving beats blocked practice
~168 questions in full course
~64 questions in full course
~81 questions in full course
Three mechanisms behind all real learning.
Retrieval practice
You learn when you pull knowledge out of memory — not when you read it one more time.
Desirable difficulties
What’s harder during study builds stronger long-term knowledge. Easier ≠ better.
Spacing effect (FSRS)
A question returns as you start to forget it — when the memory trace strengthens most.
How memory works — and how reviews make it last
Learn something once and you forget it within days. Kavke schedules short reviews exactly when you start to forget — and each one locks it in for longer. Try the sliders.
Chance you recall (%)
Days since first studying
With Kavke — after 60 days
96%
Without reviews — after 60 days
38%
Reviews in this window
3≈ 1 min total
This chart runs the same FSRS model that powers Kavke (power-law curve). Active recall and spaced reviews are among the most robustly documented learning techniques (meta-analyses: Rowland 2014; Adesope et al. 2017; Latimier et al. 2021).
An app that teaches you how to learn
Your first course after sign-up — “How we really learn” (7 lessons) — teaches exactly these mechanisms. You start by understanding how your memory works.
Everything else Kavke does.
Details for the curious — expand what interests you.
Sharing & the course catalog
Share a course in one click — as an editable copy, a preview, or a shared original. Publish it to the catalog and build a community.
Public link
Copy the link, send it, done
Three modes
Preview · copy (fork) · co-edit
Course catalog
Publish — others find and copy it
Access control
Expiry date, sign-in required
13 question types
Multiple choice, true-false with justification, open questions, fill-in-the-blanks, matching pairs, ordering, calculations, scenario-based application… AI fits the question type to the content and grades open answers semantically — asking a follow-up when an answer is incomplete. Don’t know? Tap “show the answer” or jump back to the lesson — no pretending.
Questions that aim at your weak spots.
Adaptive tests find your gaps — you don’t waste time on what you already know. Turn weaknesses into strengths.
AI narrator & voice answers
Read every lesson or listen to it. The AI narrator reads section by section and highlights the one you’re hearing; you control it from the lock screen. Answer open questions and Feynman checkpoints by voice — learning works on a walk or in the car too.
Learning a language?
Tell Kavke which language you’re learning and the questions will practise the target language instead of translating it. Six languages to choose from.
Short sessions. Results for years.
Start today — on iPhone, Android or in the browser. Your first course “How we really learn” is waiting after sign-up.
No payment card.