Tiles before writing Build a sentence with tiles on N5 and N4, then try writing it yourself once you’ve mastered the tiles. Your existing progress is preserved. The difference in context After a mistake, compare both meanings and read a short explanation of the difference in the sentence you are practicing. Spellings with examples New examples show alternative spellings of a pattern, kanji readings and usage notes. Kana remains the main spelling where recommended. Meaning in writing Feedback after a writing attempt distinguishes meanings of the same form, such as appearance and hearsay with そう.

Statistics only with your consent You can help improve Kaname by enabling anonymous usage statistics. No identifier or sentence content; withdraw consent in Settings. The guide is close at hand The guide opens from Settings. A book button on Today opens it too. Sibling status updates automatically The review marker in Your Japanese changes when reviews become due, including while the screen stays open. Choose where a sentence appears Tapping “Where should it appear” opens a separate list: examples only, or examples and practice. AI access during the grace period When the App Store confirms a renewal grace period, AI remains available until it ends. Purchase status refreshes after the network returns, without restarting.

21 new contrast pairs From N3 to N1: のみ or だけ, つつも or ながらも, ためには or ために and more. No more guessing Questions where both answers were correct and only a shade of meaning differed now have one right answer. A few pairs that can’t be asked that way are gone. みたい has two cards “Probably” (風邪みたい) and “like” (魚みたい) are two meanings, so they are now separate. The gap doesn’t give away the kanji When only the right answer had kanji, you could pick it without knowing the form. Now the options look alike. One number for “Today” The badge on the tab and on the app icon is reviews plus new points, the same as the buttons. The reminder says the same and shows your streak. Weakest spots “Needs work” marks only your three weakest points, not half the list. An example to try The sentence editor offers three ready sentences with translations. Tap one to insert it.

Bars on a grammar point count what the catalog counts The catalog percentage could be lower than every bar on the grammar point’s card. Now the bars show the same mastery, and “Overall” at the top is exactly the catalog number. The AI button waits for your quota If the app can’t check how many requests are left, “Check with AI” is grayed out and “Try again” stands next to it. The keyboard gets out of the way In a question and in the sentence editor, tap the background or drag to hide it and reach the buttons underneath. Deleting a sentence asks first In “My sentences”, swipe a row to reveal “Delete” and confirm before it goes. The same button is on the sentence screen. Sentence search is always in view The field sits above “My sentences” and also searches the grammar point assigned to a sentence — its spelling, reading and meaning.

“To fix” shows whole words The list under the AI verdict used to cut the correction mid-word. Now it shows the pair the AI pointed at: 泳ぎています → 泳いでいます. Kana that sounds different no longer passes as “Correct” しかん for じかん is a different word. If the AI lets it through, the app compares your spelling with the reading of the model sentence and tells you what sounds different. The gap covers the construction; the verb stays in the sentence Only the correct answer carried a verb, so it could be picked without knowing the form. Now you choose the construction itself: 忘れない＿＿ → ように. Similar meanings now say how they differ ようとする and てみる no longer both read “to try”. The difference now stands in the answers themselves. もう has two cards “Already done” (もう食べました) and “by now” (もう帰ってもいいですか) are two senses. The meaning question no longer has an answer that does not fit the sentence.

“I don’t know” instead of guessing There is now a second button under “Check”. It counts as a mistake and shows the correct answer — but it does not make you guess when you simply do not know. The session summary counts admissions separately: that is different information from “how often I guessed wrong”. After a mistake, the comparison is about the two you actually picked Explanations are often written to set a construction apart from a third one — and that third one showed up on screen instead of the one you chose. Now only what describes this pair remains. Two spellings of the same word no longer compete くらい and ぐらい are one particle in two uses. They no longer appear as each other’s wrong answer, because in such a sentence both forms are correct. The number on “Today” counts constructions, not questions Two new constructions showed a six, because each one carries several questions. The badge now counts the same thing as the button below it. The translation can be hidden again The button stays on screen after you reveal it, so you can cover the sentence and try again without the hint.

After a mistake you can see how the two grammar points differ If the explanation of either one speaks about the difference from the other, that sentence now stands next to the question you just got wrong — instead of two taps away on the card. The catalog already carried over two hundred such sentences; only the way onto the screen was missing. The card leads to the grammar point it mentions The explanation said "against ために the difference is one of register", and you had to find that card yourself — by a spelling you often cannot type. It now sits under the explanation, together with its meaning. A second attempt counts as a hit When the option you picked really does describe the sentence shown, and only answers a different question than the one asked, a second attempt in the same session goes to your progress — not just to a message on the screen. Options stop meaning the same thing in translation Five grammar points led their translation with a bare function word, so options that could not be told apart ended up side by side. The translations now say what the grammar point does. For the twenty-eight pairs that are still easy to mix up, a sentence about the difference itself was added. Settings fit on a narrower phone The language choice showed "Matc…ystem" where there was room for the whole thing. The value no longer shrinks at the cost of its own meaning.

Levels no longer wait on one or two grammar points Moving to the next level used to depend on a number that only grows after eight questions of each type, so a point you knew but were rarely asked about could hold the whole level back. The gate now asks how you did on what was actually tested. Material from the next level arrives once nothing genuinely needs work. The bars say they show answer strength The bars on a grammar point show the strength of your answers in a given question type — and that is what they are now called. The number in the catalog is different and lower, because it multiplies strength by coverage. Both used to be called the same thing: percent. You can see what the catalog percentage means Under the first level there is a sentence explaining where the number on a grammar point comes from: strength multiplied by coverage, that is by the number of questions you have been asked. Coverage is full after eight questions of each type, so early on the number stays low even when your answers are right. VoiceOver now names both quantities too. “Nothing left to add” instead of “more tomorrow” On the day the catalog runs out, the new-grammar button used to promise more tomorrow when nothing was coming. It now says that only when there is something to come back for.

You can see what went well and what to fix The AI check has always judged three things separately: whether the Japanese itself is correct, whether the pattern you were practicing was used, and whether it sounds right. Only one headline reached the screen. Now the result is laid out as two lists — what you did well and what needs fixing — including when nothing was found. A pattern used well no longer falls over a slip A sentence where the practiced pattern is right and the mistake is elsewhere now gets “Almost” instead of “Not this time”. The card no longer drops to the bottom of your reviews and comes back in ten minutes — it returns in its own time, just without moving up. The correction shows what actually reads differently In the corrected version, what is highlighted is the part that reads differently from what you wrote. Added kanji, a comma or a full stop stay gray, because writing a word in kana is never a mistake — until now both kinds of change looked the same and you had to guess which one mattered. You can disagree with “Correct” too The “I disagree” button now sits beside every AI verdict, not only the negative ones. If the model passed a sentence you know has a mistake in it, you can say so — the answer is then recorded as a mistake rather than a success.

Answering now takes two steps Tapping a choice selects it; the answer is sent only when you tap “Check”. Tapping the same choice again clears it. Until now a single brush of the screen was an answer — and with four choices a slip of the finger looks exactly like not knowing. Typing had this button all along; now both paths end the same way. Questions about cause have one right answer In sentences with から, ので, ために and ものだから a wrong choice could be a form that also fits that very sentence — a second right answer, counted as a mistake. Such pairs are now listed on the sentences where they really do replace each other, and they no longer meet in one question. A new pattern gets its full set of questions at once After you meet a pattern, the app now asks about it three times right away instead of once and slipping the rest back later as a “review”. The review button no longer lights up a minute after you learned something new: the day closes where it started. The widget tells you when the next review is due With nothing to review, the largest text on the tile was a dash — it said nothing. Now it shows the time of the next review, and the caption says what that time means. When there is nothing to schedule, the space stays empty instead of faking a number. The buttons under an answer match On the sentence feedback screen “My sentence was fine” stood next to “That's right” — two labels of very different length, so one button was twice as tall as the other. The pair now reads “I disagree” and “I agree”, and both buttons share one height whatever the label says. The dashboard shows how long until the next review The review button showed only a time; the countdown (“in 3 hours”) lived on the pattern card. Both sides of the app now say the same thing about the same moment. For tomorrow there is no countdown — it would repeat what the date already says.

The large tile shows your whole family of apps The review widget now comes in a second size. The small one looks exactly as it did — the number of reviews and the time of the next one. The large one adds what the “Your Japanese” section shows: a row for every app in the family you have on this phone, with its progress and a marker for overdue reviews. Rows you hid in Settings stay hidden here too. Links from another app land on the construction When another app in the family points you to a construction you struggled with, Kaname now opens its card directly, instead of the home screen where you would have to find it yourself. The card opens even when the construction belongs to a level you have not bought: the explanation, the formation and the examples are there to read, and buying starts only when you tap “Unlock”. The question counter changes with the question “Question 3 of 10” used to jump at the moment the result appeared — above the question that was still on screen. It now changes as the next question comes in. It also counts questions only: the screen introducing a new construction no longer adds to the count, because there is no way to answer it wrong. VoiceOver no longer reads your language with a Japanese voice Prompts in your review history and the titles of construction pairs were handed to VoiceOver as Japanese, so it pronounced them with a Japanese voice. The text itself now decides: Japanese is read as Japanese, with its kana reading, and interface prose in your own language. An option no longer means the same as the right answer In meaning questions and in fill-in-the-blank questions, the answer next to the correct one could be a construction separated from it by register alone — a second correct answer, counted as a mistake. Sixteen such pairs are now recognized, and they will not meet in one question.

You can disagree with the AI If the model rejects your sentence and you think it was fine, say so with one tap. Your answer counts, and the correction and explanation stay on screen — they are worth reading either way. One thing you may be used to has changed: the grade does not enter your review schedule until you confirm it — before, it was recorded on its own, right after the model answered. Writing in kana is no longer a mistake A sentence written without kanji where the dictionary uses them will no longer be marked down. The kanji spelling now appears as an alternative, not as a correction — because writing a word in kana is never a mistake. Your own sentences, for reading only For a sentence in My Sentences you now choose whether it comes back only among the examples for its construction, or in questions as well. Until now there were two states: both at once, or neither. The choice survives restoring a backup. Fill-in questions no longer give themselves away An option that produced an ungrammatical sentence once slotted into the gap — the same particle twice, for instance — could be ruled out at a glance, without knowing any grammar. Those are gone, along with options meaning exactly the same as the correct answer. You choose which apps you see In Settings, under What you see, you decide separately about the recommended apps and about the rows in Your Japanese. When something is hidden, the section says outright that it is showing an incomplete picture — what you have done does not change, and a hidden row comes back at any time.

The family section now counts ten apps Keigo (politeness), Onomatope (mimetics), Kifuku (pitch accent) and Shindan (level test) have joined the family. On the Today tab, in “Your Japanese”, they stand at the end of the list, in release order – the same order holds in every app of the family, so the section reads as one picture no matter which side you look at it from. Four new cards in “More Japanese Practice” In Settings each of the four new apps has a card of its own, and it leads straight to that app's App Store page rather than to a search. A sister you already have installed opens right away, without a detour through the store. Progress from the four new sisters finally shows up Kaname looked for progress only from the apps it knew by name, so Keigo, Onomatope, Kifuku and Shindan had no way of appearing in the section – it looked exactly as if they were not installed at all. Their row now shows the percentage and the weakest spot, just like every other sister's row.

A meaning question has one correct answer The options could include a pattern meaning exactly what the question asked about – 「〜ことができます」 (“can, be able to”) next to 可能形 (“can, be able to, be possible”), or 「から（理由）」 next to 「ゆえに」 (“because of, owing to”). Both glosses were true, so the question had no answer, and picking a good one could count as a mistake. Those pairs are now recognized and never stand together. Options still come from related patterns – that is what makes this question worth asking – but related is not the same as interchangeable. A sentence written in kanji where the pattern uses kana now counts In writing tasks the pattern 「もう少し時間がほしいです。」 keeps 「ほしい」 in kana, because that is how the catalog stores it – while 「欲しい」 is just as correct and just as natural. Until now such a sentence went to self-assessment, so the app asked you about something it already knew. A matching reading is now enough, in both directions. Your sentence stays on screen exactly as you wrote it – it is never swapped for the pattern. Romaji is still a different matter: there, spelling is judged by you or by the AI check. The percentage at the end of a session counts correct answers Under “Correct 8 of 8” stood the average mastery score – a different quantity, with a percent sign, three rows below. Both numbers were true, but it read as “eight out of eight, so 99%”, and a session with even one writing task could never reach a hundred, because self-assessed “good” has a ceiling of its own. That row is now accuracy: eight out of eight is one hundred percent. The mastery score has not disappeared – it stayed where it means “how much of this you know”, on the pattern's own screen. The highlight now appears in both places on the check screen The sentence shows up twice on that screen: in the question card and, below it, in the card with the translation. The pattern being practiced was marked only in the first one, so two identical sentences one under the other differed in color for no reason and looked like a glitch. Both now point at the same place. The progress bar fills up on the last question At “7 / 7” the bar sat at six sevenths – exactly where it sat at “6 / 7”, because the label and the bar were measuring two different things. The bar now shows what the label says. The price is a bar that is no longer empty on the first question, and that is a deliberate choice: a bar is not a counter, and one seventh reads as “you are near the start”. The analysis counter tells the truth about when it resets A subscription's analysis allowance renews from the day you bought it, not on the first of the month – bought on the 20th, it runs from the 20th to the 20th. The screens said “this month” next to it, which simply was not true, even though the renewal date beside it was right. They now speak about the period. The free allowance still renews on the first of the month, and the wording for it is unchanged.

Options in a gap differ by form, not by vocabulary Until this version, a wrong option in a gap question came from another pattern's example sentence — and the verb of that sentence came with it. A sentence about reading a book would get 「飲みません」 among the answers, so you could rule it out with a dictionary alone, knowing nothing about the form being practiced. Measured across the catalog: 1,112 of 1,438 questions had at least one such option, and 168 could be solved outright without any grammar. Every option now carries this sentence's vocabulary — only the form differs, which is exactly what the question asks about. The pattern in question is marked in the sentence In a “what does this mean here” question the prompt named the pattern, but finding it in the sentence was on you — even though the answers are meanings and test nothing of the sort. The practiced fragment is now marked in your accent color, the one you picked in Settings, with an underline added when the system “Differentiate Without Color” setting is on. Readings above kanji stay where they are. Gap and contrast questions get no marking, because the spot is already visible; writing tasks get none by design — marking would hand over the length and place of the change to someone who has yet to write it. The dot and the count now mean “there is work”, not just “there are reviews” The marker on the app's row and the count on the Today tab asked about reviews only. So the first day of study looked exactly like a finished one: no reviews, because there is nothing to review yet, while a full session of new patterns waited on the Today screen. Both now count the whole day's plan. The Review button stays with reviews alone — it says what it will unlock, not whether there is a reason to go in. The “Got it, ask me” button spans the full width After a new pattern is introduced, the button that moves you on was a narrow pill in the middle of the bar, while the other two buttons in the same place in a session — Next and Done — span the full width. It now looks like them.

An answer that reads the sentence correctly no longer gets “Not this time” In a question about meaning, one of the options can be a pattern that contains the one being asked about — 「ても構わない」 next to 「ても」, for instance. Its meaning is a true reading of the sentence on screen, so choosing it was sound reasoning, and the app counted it as a miss. It now has an outcome of its own: it says the sentence was read correctly, and shows what the question was about. It counts as neither a hit nor an error — it leaves mastery alone, does not move the review date, and stays out of your accuracy. The question simply comes back. Contrast questions stop coming back with the same pair A hundred and sixty-eight patterns out of a hundred and eighty-eight have exactly one contrast pair, and nearly every pair has two sentences. Question choice followed your weakest skill, so while contrast sat low it kept returning to those same two screens. The weight of that question type now fades after each pass through all the material there is. Contrast questions do not disappear: they get rarer and spread out over time. The same pair no longer asks twice in one session A session used to make sure the same screen did not come back. That was not enough, because what is being practiced is the distinction, not the sentence: 「は」 against 「が」 in a second sentence is still 「は」 against 「が」. When both halves of a pair fell due on the same day, the same distinction could be asked twice in a row. Answer buttons of equal height Height followed the length of the option, so an answer wrapping onto two lines made a tile twice as tall as its neighbor. Size means nothing here, and it looked as though it did. Tiles in one question now match the tallest, while still growing with the text at larger type sizes rather than clipping it. An announced level and an untouched level stop looking alike An empty progress bar meant two different things at once: “there is no material yet” and “the material is there, nobody has touched it”. The first has no scale on which zero would say anything; the second does. Review intervals written the way the language writes them The “Interval” row on a pattern's page showed a fraction of a day with a point that ignored the language of the interface. Every other number in the app followed the local convention. Now this one does too. A faster start The Today screen recomputed a fingerprint of the whole catalog more than a hundred times per launch instead of once per pass — five hundred and sixty-three patterns each time. After the fix: eight passes instead of a hundred and nine to two hundred and seventeen.

A sentence from the example no longer gets “Not this time” With 〜をおねがいします the prompt demanded a construction the model answer could not contain: を does not follow まで, and a note in the very same entry said so. Type the sentence from the example and you were told you were wrong, with the screen then suggesting the Japanese it warns you against. The example comes back as a fill-in-the-gap task — where it teaches exactly the thing that を drops out. The review button no longer stays off once a review is due Left in the background, the app aged along with the number of reviews waiting: the dot on the icon could show a figure while the button underneath sat dark. Anyone who trusted the button missed the review and had no way of knowing it was a fault. The Today screen now watches the clock the same way the badge does — and wakes by itself the moment a review comes due. Without a connection, AI analysis says so All the learning works offline: the material is in the app, your progress is on the device, the phone schedules the reviews. Only AI analysis needs a connection — and its button looked the same in every condition, failing only once you pressed it, like a fault. It goes dark now, with a line underneath saying why. A service switched off and an allowance that could not be read get their own sentences, because those are three different things. Similar patterns now show up in sentences, not just in prose A pair of patterns used to be a title and one paragraph. The material had shipped from the start and sat unused: a hundred and seven pairs, each with at least two sentences where one form is natural and the other is not — with the context and the reason. The only way to reach those sentences was to get an exercise wrong. Now you open the pair from the pattern's page and see them side by side.

KANJI IN A JAPANESE TYPEFACE, NOT A CHINESE ONE On a phone with an English or Polish interface the system substituted a Chinese typeface, and some kanji are shaped differently in it. Every Japanese line now tells the system what language it is in — and looks the way it looks in Japan. YOU CAN SEE HOW MUCH OF THE SESSION IS LEFT The progress bar runs the full width below the navigation bar, instead of sitting in the title at the width of a thumb. In a 25-question session one step used to be thinner than a hairline. VoiceOver reads it as "Question 7 of 25". ANSWER CHOICES ARE VISIBLE IN THE LIGHT THEME Answer tiles and the typing field were white cards on a white screen, held together by nothing but a hairline border. The question screen now has a gray background, and the cards look like cards. YOU CAN TELL WHY THE ICON HAS NO NUMBER If badges are turned off for Kaname in the system Settings, the app's Settings now say so and lead you to where you can turn them back on. Before, the number was simply missing with no way to guess why.

FASTER SWITCHING BETWEEN TABS The Today screen recalculated everything each time you came back to it, and Levels did the same on every redraw. They now do the work once, and only when something has actually changed. THE AI ANALYSIS SCREEN SAYS WHAT IS GOING ON When your allowance or the offer cannot be loaded, the screen says so and offers a Try again button. A failed purchase now gets a message instead of quietly returning to the button, and a purchase waiting for a parent's approval says that it is waiting. THE SENTENCE YOU TYPED IS NO LONGER LOST If saving fails, the editor stays open and tells you why, instead of closing exactly the way it does after a successful save. A LEVEL WITH NO REVIEWS YET SAYS NOT PRACTICED Instead of a coloured bar at zero that looked like a result. The catalog has done this for a while; now the dashboard says the same.

A SUBSCRIPTION YOU BUY NOW WORKS RIGHT AWAY After a purchase the app confirms your subscription with the server immediately instead of waiting for a restart. Restore Purchases does the same. A failed restore can be dismissed, and closing the Apple ID prompt is no longer reported as an error. YOU CAN SEE WHETHER A SUBSCRIPTION RENEWS OR ENDS The screen said it renews even when renewal was turned off. It now says it ends and warns that AI analysis will stop working after that date. A failed payment gets its own wording and its own advice. SCANNING ASKS BEFORE IT REPLACES YOUR SENTENCE Text from a photo went into the field without asking, even when your own sentence was already there. Grammar recognition now says what it did: filled empty fields, replaced them with your consent, or changed nothing. SUBSCRIPTION PERIODS IN THE RIGHT LANGUAGE The offer tiles took the period wording from your phone language instead of the language the app is set to. YOU CAN SEE THE APP ASKING ABOUT YOUR ALLOWANCE The AI analysis row in Settings went blank while the request was in flight if you had a subscription — no number and no sign that anything was happening.

N2 AND N1 — THE WHOLE JLPT RANGE The last two levels are here. Anyone who bought Premium earlier gets them at no extra cost. Levels can also be bought separately or in bundles — buying at once is always cheaper than piece by piece. SCAN JAPANESE FROM A PHOTO Take a photo of a page or pick one from your library. The text is recognized on your phone; the photo is never sent or saved anywhere. It reads vertical writing too, so a manga speech bubble can be put together from its columns. AI ANALYSIS Write your own sentence and the model tells you what exactly is off and why. It also spots patterns in text you paste in. A separate, optional layer — without it every number and every due date comes out the same. A few calls a month are free. SMALL THINGS YOU COULD SEE • A correctly written sentence checked by AI finally scores a hundred percent, not seventy- nine. • The verb form check works without AI too — a malformed て or た form no longer passes as correct. • Picking a sentence after a scan looks like buttons, because that is what it is. • Catalog search folds Polish diacritics: “okreslenie” now finds “określenie”.

The same questions stopped coming back in one session You reported it three times and you were right every time — the cause was different every time. The last one: a contrast exercise belongs to the pair, while the record of what had been asked was kept per grammar point, so the same task could come up once "for は" and once "for が". Measured before the fix: 26 sessions out of 60. After: none. The last gap is closed too — the one that showed the very same sentence again once the material ran out. Questions stopped running ahead of you A contrast question — "how do these two differ" — only makes sense once you recognize the first of them. The engine held that line when picking a question type, but not when it dropped down looking for material, so the third question about a brand-new grammar point let contrast back in through the side door. Measured: almost one question in four bypassed the threshold. Now none. The dashboard and the session stopped stalling Choosing an answer took 170 milliseconds; it now takes twelve. The dashboard worked out the whole session plan eighteen times per redraw, and the catalog was rebuilt on every keystroke in search. None of it looked like a bug — it looked like the app dragging its feet. A grayed-out button says why it is grayed out "Review" and "New grammar points" went gray without explaining themselves — and they go gray for different reasons, each meaning something else. Each now carries a short reason on the same line: when the button will light up, whether today's limit is used up, or whether the material has run out. A finished day means the learning goal Today's new grammar points close the day, and nothing else does. Until now closing it also required an empty review schedule — and a single review falling due in the evening held the day open when there was nothing left to close it with. Reviews have their own marker and their own path. The daily limit is a ceiling, and you can see what is left The number on "New grammar points" is what remains of today's limit, not the length of the session: with one of five introduced, the button reads "· 4". Once the limit is used up it goes gray and says so, instead of quietly adding material beyond what you set. Starting from a grammar point's page asks first Opening a grammar point from the list and starting it there went over the daily limit without a word. It now asks — once, with the number — and goes ahead only after you confirm. Material above your stage works the same way: the button does not go gray, it warns and asks. "I already know this" silences a grammar point A grammar point you know from elsewhere does not have to be walked through question by question. Marking it writes in full mastery and takes it out of the queue — as a declaration, which is what it is, so nothing pretends it was practiced. Weak spots are drawn by weight, not by order The weak-spot basket sorted until it was done and kept showing the same five. It now draws from weights, so a weak spot from the second ten gets its turn as well. The threshold at which a grammar point lands in that basket came down from 45 to 40 — the same number the color scale paints by.

Review count widget The Home Screen and the Lock Screen now show how many questions are waiting for review — and when none are, the hour from which they will be. The number comes from the same calculation as the “Review” button inside the app, so it cannot promise something that is not there once you open it. The week chart is about work, and the day section about your stage Tomorrow’s bar showed only the due dates already stored — but most of tomorrow’s reviews are created by today’s session, because the intervals start at a few hours. It understated the day by dozens of questions, and the next morning came as a surprise. It now adds what is certain: material studied today that returns tomorrow, and the points introduced along the way. Three new points a day instead of five The default pace drops from five to three. Five turned out to be too many, because every new point comes back several times and adds to the reviews for a whole week. If you have set your own pace, it stays exactly as it was — the change applies to new installs only. Sentence translation after you answer The check screen now shows the translation of the Japanese sentence. On tap by default, adjustable in Settings next to furigana. It is not shown with the question itself and will not be: almost half of the translations contain the very word that is the correct answer. After a mistake you see the difference, and each grade shows when the question returns Until now a wrong answer showed you what was right, but not how it differed from what you picked. Both now stand side by side, each with the point it belongs to — and where a contrast pair exists between those two points, its sentence about the difference comes with them. The four grades in a production task decide when the material comes back and said so by name only; each now carries the actual interval, so the choice stops being guesswork. Statistics with a 7, 30 or 90 day window The statistics screen showed a week of answers and a single number for the whole history — and “87%” does not tell you whether things are getting better. You now pick the period at the top, and below it stand accuracy day by day, answers per day, and the grammar points finished along the way. A day without study has no point on the chart: no answers is not the same as nothing but mistakes. “Your Japanese” shows where the reviews are Rows for the other apps in the family told you how the learning was going there. Now you can also see which of them has something to review right now. Content review: nine constructions split apart A construction carrying several roles under one spelling could not be learned — the question “what does it mean” had two correct answers. が, と, ところ, ために, によって, きり, だって, という and ということだ now stand as separate entries; ところ alone became three, because “about to leave”, “in the middle of leaving” and “just left” are three different things. Five entries that are names of forms are no longer asked what they mean. The whole catalog went through a sentence-by-sentence review in both languages, and pairs whose difference is one of tone got a new exercise: two correct sentences and a question about which one carries the role. A particle typed in romaji now matches itself A sentence typed in the Latin alphabet was told that “your answer does not show the grammar point you are practicing”, even when it was perfectly fine: romaji writes the pronunciation (“wa”) while the catalog writes the spelling (は). This affected は, へ and both を entries — the particles typed most often.

The next review time tells the truth Once today's review allowance was used up, the chart showed zero for today and marked the day as done — while the button below promised a review later the same afternoon. Both numbers were true on their own and contradicted each other together, because they counted different things. The button now answers the same question as the bar above it: when there will genuinely be something to review. Weak spots lead to the grammar point Rows in that section on the Today screen had an arrow, but tapping them did nothing at all. They now open the grammar point — its four competences, the next review date and everything else the row is talking about. The whole row responds to a tap On several lists a tap only worked on part of the row, and on Japanese expressions it was lost entirely. It looked like the app stuttering, and that is how it was reported. A row now responds across its whole surface, first time — in Grammar, in the study plan, in Sentences and in Settings. The pronunciation button is easier to hit as well. Readings no longer carry over to the next question With readings set to reveal on tap, the next question about the same grammar point could start with the furigana already showing — a hint nobody asked for. A second attempt after a mistake still keeps it, because there the reveal was a deliberate choice.

No more of the same sentence twice A fill-in-the-blank question and a contrast-pair question could show the same sentence with the blank in the same place — once with four options, once with two. It looked like a mistake by the app, and fairly so. A session now watches what appears on screen, not only what the question was built from. A reminder when reviews are waiting Turn it on in Settings, along with the hour. It arrives once a day, and only when something is genuinely waiting — a day with no reviews stays quiet. In holiday mode no reminders arrive at all. Permission is asked for only when you switch this on. The Japanese keyboard, straight away A new setting: the field for your own sentence can open on the Japanese keyboard rather than the one you used last. Off by default, because romaji counts just as it always has — turn it on if you write in kana anyway. A second attempt without a peek After a wrong answer in a multiple-choice question you get another go — but the correct answer and the explanation stood above that button, so there was nothing left to test. They now sit behind "Show the answer", and the screen keeps only the sentence from the question. Two glosses now say what they mean The translation of 「と」 now leads with a plain "and". It used to speak of a complete list and of company, but not of what と means most of the time. 〜ています in its resulting-state sense now reads "lives, is married" instead of "is in the state after…". Both were reported by a user.