Want in? Closed beta is in Discord
If you want to join the closed beta and get your hands on Kumi early, hop into our Discord: https://discord.gg/learnkumi
Paper’s still warm
Week one of beta and I feel like we’re watching a thousand tiny folds happen in real time. You all poke at the product, we sprint back in with tape, and somehow it’s turning into something sturdier every hour.
Also, quick context: we brought on an editor. That’s why we’ve been a little quieter on socials. We’re not disappearing. We’re loading the slingshot. Content is about to ramp.
Week 1 of beta: what changed that actually matters
TL;DR
Imports are less brittle (and way less scary).
Reading support got a serious upgrade (furigana options).
The dashboard + Daily Fold are more honest and more responsive.
Input got smoother (faster IME, better in-lesson typing, keyboard shortcuts).
1) Imports: fewer bottlenecks, less pain
JLPT self-report and Anki import aren’t chained together anymore. You can do one without waiting on the other.
Anki import got a full glow-up: higher upload limit, faster runs, and a bunch of parsing fixes that hit real-world decks.
If something fails, it should fail like a polite product now. Not like a JSON horror story.
2) Furigana is here (and it’s not all-or-nothing)
You can turn readings on in three modes:
Off (raw mode)
Adaptive (shows furigana on kanji you haven’t mastered yet, then fades it out)
Always
And there’s a slider so you can pick how much help you want. Plus a quick toggle inside lessons so you don’t have to menu-dive.
Adaptive, hiding furigana for word a user has learned.
Always, showing furigana regardless if a user has seen it previously.
3) Daily Fold + dashboard with more momentum
The Daily Fold counter updates immediately after you hit Continue.
Effort counts now. You try, you get progress credit. (Your wrong answers still get tracked for smarter reviews later.)
Post-review navigation is clearer and the labels actually match where you’re going.
4) Lessons feel smoother (especially input)
IME inside lessons is faster and more reliable.
Manual mode can advance with Enter.
Long sentence tile-assembly wraps cleanly.
Meaning checks are smarter: we can recognize contextually-correct answers even when they are not text-perfect, and give helpful “close but not quite” feedback when you are lukewarm.
Multiple choice can be answered with 1, 2, 3, 4.
5) IME input got serious (it now feels like a real IME)
Romaji becomes kana as you type, and stays dotted-underlined while it’s unconfirmed.
Hit Space to convert kana to kanji. Hit Space again to cycle candidates. Hit Enter to commit.
F6 switches the current segment to hiragana. F7 switches it to katakana.
Mixed-script words behave properly now (like ムッと, パンや, ガス代).
There’s a subtle あ/ア toggle for mixed-script segments so you can flip scripts mid-word (and your sokuon ッ stays in the right script).
Tiny thank-you corner (beta brain = precious)
If you posted feedback this week. Seriously, thank you. This is the fun part of beta. The part where we find the sharp edges fast and sand them down while the paper is still flexible.
That’s it for this week.
What did you try in the beta that surprised you? In a good way or a “lol what” way?
— The Kumi Team