A lot of what people study now isn't text on a page. It's a YouTube lecture. A podcast at 1.5x. A 90-minute recorded seminar that nobody is going to rewatch.
v1 of Klarrity couldn't see any of that.
That's the gap v2 closes. But the more honest framing — and the one I want to walk through here — is that v2 didn't invent a new loop. It took the three-step loop that v1 already worked, and strengthened each step.
v1 in one sentence
Klarrity is a Chrome extension that turns whatever you're reading into flashcards. You highlight a passage on a page — an article, a research paper, a lecture handout. Klarrity reads it and gives you back well-formed Q&A cards. From there you export to wherever you actually study: Anki, Quizlet, Notion, Obsidian, or a plain CSV.
The point of v1 was to close a gap nobody else was closing. Flashcard apps live in their own walled gardens. The content you actually want to learn from lives in your browser. So most students end up retyping definitions into Quizlet, or copy-pasting paragraphs into ChatGPT and asking for cards. v1 made that whole detour disappear.
Highlight, generate, export. One loop, in the place you were already studying.
That's the loop. Three steps. v2 didn't change it.
Step 1: input
v1's input surface was a webpage. That's a fine surface. It's also too small.
Most modern study sessions don't happen on a page. They happen on YouTube. They happen on a podcast app. They happen on a recorded Zoom that someone uploaded to a course portal.
Study Klips extends the loop to video. You're watching a lecture. You hit Klarrity, choose a window — last 30 seconds, last 60, last 2 minutes — and the extension pulls the transcript and turns just that slice into a deck. Same gesture as v1. Same outcome. The surface is just bigger now.
This is the hardest piece of v2 to build, and the one I'm proudest of, because the temptation is to do the whole video. Don't do the whole video. The thing students actually want is "wait, what was that last bit?" — they want a deck on the moment they didn't follow. v1's input surface was the page. v2's is the page and anything playing in it.
Step 2: generation
v1 generated cards from whatever you fed it. The flaw was silent: you'd get a 40-card deck on a topic where you only needed five, because the engine doesn't know what you already know.
Three v2 features sit on this step.
Klarrify: diagnostic-first
The most opinionated thing in v2.
The best deck is the smallest one that still teaches you the thing.
Klarrify quizzes you first. Five MCQs in three minutes. Then it only builds cards on the concepts you missed. You skip what you already know. The decks come out half the size and twice as useful.
Most flashcard tools quietly assume more cards = better learning. They don't. The single biggest waste in flashcard practice is reviewing things you've already mastered. Klarrify removes that waste at generation time, before you ever open Anki.
Cloze deletion
v1 was Q→A only. Cloze deletion — fill-in-the-blank — is what serious Anki users actually want, and v1 was missing it. Same generation engine, more flexible cards. Fixed.
Streaming generation
v1 made you stare at a blank popup for 10 seconds while a deck generated. v2 streams the cards out as they're written. Same call to the model under the hood. Subjectively makes the loop feel three times faster. You stop closing the tab.
Tiny change. Completely changes how often you actually use it.
Step 3: export
v1 already did this well. Anki, Quizlet, Notion, Obsidian, CSV. The cards belong to you.
The one rough edge was Anki: download a .apkg, open Anki, drag the file in, click import, dismiss the dialog. Five steps for what should be one.
AnkiConnect direct send removes the manual step. One click in Klarrity → cards land in your Anki deck. Same export targets, less friction. Small piece of plumbing, completely changes the loop.
The export-everywhere stance is the load-bearing piece here. Flashcard apps that lock you in are the reason this product had to exist. I'm not going to recreate that problem one layer up.
The pattern
Every v2 feature targets a specific weakness in a specific step of v1's loop.
- Step 1 weakness: input limited to text on a page. → Study Klips for video.
- Step 2 weakness: decks too big, dead-air wait, only one card type. → Klarrify, streaming, cloze.
- Step 3 weakness: Anki import was clunky. → AnkiConnect direct send.
Nothing was added because it'd be cool to have. The pitch is unchanged ("highlight to flashcards") — v2 just made the pitch true in more places, with smaller decks, faster, with less manual work.
There's also a layer of polish that doesn't fit the loop framing: dark mode, a new geometric logo, a ground-up popup redesign. These don't change what Klarrity does. They change how it feels to use it for the fourth hour at 1am. That matters more than I expected before I shipped them.
On pricing
v2 moved to $7/mo or $59/yr. v1 was $5/mo.
Existing users on the old $5/mo plan stay on $5/mo. Forever.
The why on the price move: v2 does meaningfully more work per user. Supadata transcripts cost real money. Klarrify is two model calls instead of one. The economics of v1 don't carry the costs of v2. The price had to move.
The why on the grandfather: existing users signed up for v1's economics. The right thing — and frankly the only thing that doesn't burn long-term trust — is to honor that. Quizlet quietly migrated their users to higher prices a few years back and the trust never came back. I'm not running that play.
If you signed up for v1, you signed up for v1. v2's prices apply to v2's signups.
It's a small product. Trust compounds. Burning it for a few extra dollars a month is a bad trade.
Stack
Built solo. Stack is small on purpose:
- Next.js for the marketing site and the dashboard
- Supabase for auth, postgres, and storage
- Stripe for payments
- OpenAI for card generation
- Supadata for YouTube transcripts (the only piece I didn't write)
That's the whole stack. One person, one repo, real users paying. v2 was shipped on a one-person budget.
Try it
If you study from things you read or watch in your browser, give v2 a try → klarrity.app
If v1 was flashcards from the web, v2 is the same sentence with two words underlined: from anywhere you read or watch, with cards focused on what you don't already know.
That's the whole thing.
See Q in action
AI agent orchestration with governance, trust scoring, and specialist agents.