UmaKuma Updates

An arrow on the right of a word's row, shown only while that row is hovered or the link has focus, opening the word on jisho.org in a new tab. Out of the way of reading and still reachable by keyboard. Requested 2026-09-04 by John.

Area

study

Name

「へんじ」 · 返事 Henji · the fuller answer, waiting one tab away

The kanji whose page you are on was dropped from each word's chips, so growth business drew three chips under a four-character word and a reader counting them finds the maths wrong. It is drawn now, ringed as the one you are on, with no href - it still leads nowhere and says so. Reported 2026-09-04 by John.

Area

study

Name

「ふぞく」 · 付属 Fuzoku · every kanji belonging to the word, the word's own included

History and Stats sit under their own Progress group, but both are your WaniKani progress and belong with the studying that produced them. Move the pages for real - /users/<name>/study/history and /users/<name>/study/stats - and give Study a second row of three: Reviews, History, Stats. Progress goes away as a group. No redirects; update every caller. Requested 2026-09-04 by John.

Area

study

Name

「いっしょのすみか」 · 一緒の住処 Issho no Sumika · the records live where the studying happens

The control over what an item chip says offered Off, kana and English - one question at a time, which was the point of it. The objection to printing both was never to seeing the pair though, it was to having no say in it, so Both is back as a fourth choice: a chip reads しゅく / Lodging for anyone who asks for that, and the other three modes are untouched for anyone who does not. It draws through the same function the hover title has always used, so what you see when you ask for the pair and what you get on a hover cannot drift into two formats, and it falls back the way the other modes do - to whichever half the chip actually has - rather than printing a joiner with nothing on one side of it. Requested 2026-09-03 by John.

Area

study

Name

「まとめてならべて」 · まとめて並べて Matomete Narabete · the reading and the English together, when you ask for them

Written with holds radicals only, so it is the same kind of fact as the Built from group in Related and should sit next to it rather than in its own block further up. Requested 2026-09-03 by John.

Area

study

Name

「ほんやくなしひかく」 · 翻訳なし比較 Honyaku Nashi Hikaku · two books on one question, side by side

A worksheet for a single character is titled Writing practice, Chosen characters, which is neither. It should name what it is showing, mentioning the kanji. Work out what every other way in should be titled too, not just this one. Reported 2026-09-03 by John.

Area

study

Name

「へいぜんとなざし」 · 平然と名指し Heizen to Nazashi · the sheet plainly named after what is on it

The characters printed on a worksheet are not links. They should go to the character's own page. Reported 2026-09-03 by John.

Area

study

Name

「ひらいたでぐち」 · 開いた出口 Hiraita Deguchi · a way out of every row

A worksheet row carries the readings; it has room for more of what is known about that character. Decide what earns the space. Requested 2026-09-03 by John.

Area

study

Name

「はやいみたて」 · 早い見立て Hayai Mitate · place the character at a glance

John: it looks really good but can someone guess what each thing does? Why is 10 strokes a different badge from the Worksheet button - do those belong together? Question the design and the layout, ask why each thing is where it is. Concise is good, not too busy. One idea of his: the 10 strokes label folds into the Show one stroke button, and pressing that is what reveals the ten numbered buttons. Requested 2026-09-03 by John.

Area

study

Name

「のぞいたさわぎ」 · 除いた騒ぎ Nozoita Sawagi · the noise taken out of the card

The subject page sections have titles now and each title already has a RESTful path behind it. Offer the link: on hover a section title becomes a link to its own view, for example /kanji/X/related, and the main kanji page can be navigated from. Requested 2026-09-03 by John.

Area

study

Name

「ねらいどおりつなぐ」 · 狙い通り繋ぐ Nerai Doori Tsunagu · each title joined to its own page

Nine sizes, none of them decided by anybody. A subject page headed itself at one size on a phone and another on a desktop; five list surfaces - study history, a saved list, what you looked at recently, a page of search results, the suggestions under the search box - each spelled the same three classes out in full, and copies age: one had quietly grown a larger desktop size the others knew nothing about, and one had lost the attribute that tells a browser its text is Japanese, which is what stops Chrome offering to translate 私自身 into "myself" and a screen reader spelling the kanji out in English. A list proposal row and a note modal each used a size nothing else on the site did. A glyph is three things now and one module owns all three: the subject of a surface, measured by length so a six-character word does not run off a phone; one of many on a browsing card; one of many in a row or on a chip. The sizing helper moved to the shared folder on the way, because a shared component had been reaching into the level explorer's private lib to borrow it. A gate in the quality check fails on a fourth size typed onto a line that renders Japanese: it found two this sweep had missed and a third that landed while the sweep was being written. A surface that genuinely needs its own shape - a game tile sized to its board, a sentence in prose, the radical picker - is named on its allow list with the reason written out. Asked for 2026-09-03 by John: reuse rather than one-offs, and go through the site rather than fixing where it is noticed.

Area

study

Name

「ぬのじのものさし」 · 布地の物差し Nunoji no Monosashi · one ruler, and every piece cut to it

The Everything about X link at the top of a single-section page such as /kanji/X/related has no styling at all - it is a bare underlined link on the page. Reported 2026-09-03 by John.

Area

study

Name

「にしきのひょうし」 · 錦の表紙 Nishiki no Hyoushi · a proper cover on the page

The vertical bar radical shows on a kanji page with an empty name, and /radicals?parts=<bar> is reached from /kanji/倫. Find out whether the dictionary has no name for it or we are failing to read one. Reported 2026-09-03 by John.

Area

study

Name

「なのるかけら」 · 名乗る欠片 Nanoru Kakera · every piece says what it is called

On the practice sheet the Show readings option is always unchecked when the page opens. The other options are remembered; this one is not. Reported 2026-09-03 by John.

Area

study

Name

「てきようおぼえている」 · 適用覚えている Tekiyou Oboeteiru · the settings come back the way you left them

The stroke picker starts with nothing picked, so the panel opens showing no stroke at all. Default to the first one: there is no point in having nothing show. Reported 2026-09-03 by John.

Area

study

Name

「つけるあしば」 · 着ける足場 Tsukeru Ashiba · a footing to land on, not a sign saying choose

The item chips printed both halves at once - キ · cape, しゅく · Lodging - and nobody had chosen that: the control was on or off, and "on" meant everything the chip knew. The reading and the meaning are two different questions, and a member reading a page wants one of them at a time. It is a three-way choice now - Off, あ, EN - held once per browser and read by every chip on every page: the parts of a character, the words a kanji is used in, the related groups, the kanji of a place name, a paste being turned into a list. Segments rather than a button that cycles, because a cycling button cannot say what the other choices are and you end up pressing it until the page looks right. The half a chip does not print stays on its title, so seeing it costs a hover rather than a trip back to the control, and each mode falls back to the word the chip actually has - a radical WaniKani draws has a name and no reading, and asking for readings must not empty a whole row of them. Requested 2026-09-03 by John.

Area

study

Name

「ちゃんとよみわけ」 · ちゃんと読み分け Chanto Yomiwake · the reading and the English, told apart

On a stroke-order worksheet the leftmost square of every chart row holds the finished character, and it is drawn faint while the same square on every practice row below it is solid. One column down the page, two weights, for the same thing. Make the chart rows match the rest. Reported 2026-09-03 by John.

Area

study

Name

「たていちれつ」 · 縦一列 Tate Ichiretsu · one column, one weight

A kanji page drew the same chip at two sizes a block apart: the kanji inside each compound under Used in words came out at two thirds of the parts under Written with and the radicals under Built from. All of them are the same component. It had carried a size prop since the days when four surfaces drew four different chips, and folding those into one moved the disagreement inside the shared piece instead of settling it. The prop is gone and every pill is one size - the subject pages, the strip of what you have selected, a paste being turned into a list, the updates a copied list has waiting. Its dead half went with it: the small variant also asked for a narrower minimum width, which lost to the default in the stylesheet and had never once applied. Reported 2026-09-03 by John.

Area

study

Name

「けなみぞろい」 · 毛並み揃い Kenami Zoroi · every chip in the same coat

100 dollars CAD and 500 Japanese yen name one amount in three tokens, and the parser takes at most one token on each side of the number, so both answer with nothing. Low value on its own; worth doing when the parser is next open. Noted 2026-09-02.

Area

study

Name

「えんちょうのめいさい」 · 延長の明細 Enchou no Meisai · the line item, spelled out at length

five hundred yen and twenty thousand yen name a price as clearly as 500 yen does, and neither parses - the amount must be digits today. Depends on the number library item 44 builds, which has to read words into values anyway for the reverse direction. Romaji units belong here too: a learner types 5 man yen before they type 5万円. Noted 2026-09-02.

Area

study

Name

「うたがわないつづり」 · 疑わない綴り Utagawanai Tsuzuri · the spelling it takes at its word

Every inline kanji, radical or word across the site is drawn through the shared pill; hand-rolled chips and misplaced large cards are replaced.

Area

study

Name

「るいをよぶ」 · 類を呼ぶ Rui o Yobu · like calls to like

A search for 500 yen answers with what it converts to and never with how to say it. The amount a learner most wants from a price is 500円 and ごひゃくえん - the currency answer had no Japanese form at all when the query was already in yen, and the reading the number answer works out sat on a line the search dropdown does not draw. An answer carries its reading as its own field now, so every surface that shows an answer can say it aloud. Reported 2026-09-03 by John.

Area

study

Name

「よみあげるねだん」 · 読み上げる値段 Yomiageru Nedan · the price, read aloud

The practice sheet carries no brand at all: on screen the page has the site's navigation and nothing that says UmaKuma, and on paper - where a sheet is photocopied, handed to a child and pinned to a fridge - there is nothing at all. Every other page wears the banner. The sheet wants the brand minimally on screen, the mark and the name above the title, and more minimally still on paper: one quiet line, the name and the address, in the grey the rest of the printed sheet uses, so whoever picks the page up knows where it came from and where to get another. Requested 2026-09-03 by John.

Area

study

Name

「みせるひょうじ」 · 見せる表示 Miseru Hyouji · the sheet shows whose it is

The one-character practice sheet was a second implementation beside the practice sheet the lists print - its own page, its own component, its own row grammar (faint model leading each row, no heading line) and none of the options, so the two looked like two products. There is one sheet now: TracingSheet takes the characters it is given and a Fill the page option shares the page's rows between them - one kanji takes the whole page, two take half each - on top of whatever rows each needs of its own, the stroke chart in Stroke order mode or the trace row in Trace mode. The character page's Worksheet links into the member's practice sheet with that character picked, Stroke order on and the page filled, so it has the same header, the same Practise, Sheet and Options rows, the same square sizes and the same Print. /kanji/X/sheet, KanjiPracticeSheet and its copy are deleted. The faint-first-row grammar of 0.333.0 goes with them: the shared row is the workbook's - the finished character to copy, faint repeats to trace, blanks to write - and a variant belongs as an option on this sheet, not a fork. Requested 2026-09-03 by John.

Area

study

Name

「へだてないますめ」 · 隔てない升目 Hedatenai Masume · one sheet, no wall between a list and a lone character

A radical and a word open with an identity card - the glyph, the English name, its type and level as pills, then meanings and readings under a credit. A kanji opens with a bare line of text above the stroke order box, so the one page most people arrive on is the one that looks unfinished beside its two siblings. Parity is structural: the same card in the same slot, fed by what the kanji page already knows. Reported 2026-09-03 by John.

Area

study

Name

「ふだんのなのり」 · 普段の名乗り Fudan no Nanori · introducing itself the way the others do

Two words for two actions, and the site had them crossed. On the list cards and the list page, Print was a link that navigated to the worksheet and asked it to open the print dialog on arrival - a button called Print that changed the page - while the stroke panel's new Print went to the one-character sheet and then did nothing, sitting in a row of its own under the drawing. One rule: Worksheet is the button that goes to a sheet, and Print exists only on a sheet, where it opens the dialog - with the scope menu where there is more than one page, as the practice sheet already does. The cards and the list page lose their navigating Print; the stroke panel's button becomes Worksheet in the header slot beside the stroke count; the one-character sheet gets the same Print button the practice sheet has. The arrive-and-print flag stays, because the sheet's own Everything choice reloads into the print layout and needs it. Requested 2026-09-03 by John.

Area

study

Name

「ひとしいやくめ」 · 等しい役目 Hitoshii Yakume · the same job everywhere

The one-character sheet ruled the whole of Now You into empty squares, which asks a child to copy from a chart four inches up the page. A workbook leads the hand before it lets go: the first practice row is the character built stroke by stroke in faint, to trace over, and every row after it opens with the finished character in faint before the blanks. Nothing below the chart is highlighted - down there every mark is something to draw, and one black stroke among them would be the only one that is not. The chart at the top keeps its black stroke, since that is what says which one is new. StrokeStepGlyph grew a ghost tone for it rather than a second component. Requested 2026-09-03 by John.

Area

study

Name

「のこるうすずみ」 · 残る薄墨 Nokoru Usuzumi · the faint ink that stays on the page

A sheet for one character, at /kanji/<character>/sheet and linked from the stroke panel. The character and its readings at the top, then every stroke as its own square the way a stroke-order chart shows it - all of them, so a twenty-two stroke character gets twenty-two squares rather than the first row - and then the rest of the page ruled into fifty-six empty squares to copy it into. The list sheets print a grid across many characters, a row each; this is the other shape, one character and room to work at it. The squares are the list sheets' own: Cell, TraceGlyph and StrokeStepGlyph came out of TracingSheet into sheetCells so both draw the same workbook square rather than drifting apart. Public like the character page it hangs off, because a parent printing a sheet for a child has no reason to sign in first, and a static segment so the stroke-order section keeps its address. Requested 2026-09-03 by John.

Area

study

Name

「ねるいちまい」 · 練る一枚 Neru Ichimai · one sheet, to work the character into

Stroke order's header carried the readings and the meaning under its title - the only block on the page whose header did two jobs, now that every other card wears the same plain bar. It did that because the kanji page had no header of its own: the card was the first thing on the page, so it became one. The page names its character now, and stroke order's bar says Stroke order and the stroke count like every other card. A section drawn on its own - /kanji/X/stroke, or the panel in a modal - keeps the line, because there is no page header above it there. summaryLine moved out of the client module into lib on the way: a function exported from a 'use client' file cannot be called during a server render, which is how the page broke the first time it was wired up. Requested 2026-09-03 by John.

Area

study

Name

「ぬぐかんばん」 · 脱ぐ看板 Nugu Kanban · the sign the card no longer has to wear

Holding a stroke still drew it in colour with every earlier stroke behind it in grey, which answers where in the character it falls but not what the stroke itself is - on a twenty-two stroke character the one you asked about is a coloured mark inside a nearly finished glyph. Only hides what is already down and leaves the single stroke on the page; with Outline on it sits over the faint whole character, which is the shape of the question: this stroke, there. A finished stroke becomes 'not yet reached' rather than a fourth state, because that is already what the animation knows how to leave wound back, and the numbers follow the same rule so nothing is numbered that is not drawn. Off by default - where a stroke falls is the usual thing to want - and remembered. Requested 2026-09-03 by John.

Area

study

Name

「にぎるいっぽん」 · 握る一本 Nigiru Ippon · the one stroke, held on its own

Stroke order draws its title in a header bar across the top of its card - a rule under it, a muted ground, the stroke count on the right - and every other block put a small label inside its padded body instead, or had no title at all. Meanings, Mnemonics and Related carried no heading, so a reader scrolling a kanji page met cards that looked like different components. One header bar in SubjectBlock, the one stroke order already used, and a title on every block that is a section of the page - taken from the same sectionTitles the section pages are named by, so a card and its own address agree. The right-hand slot is that panel's too: where its stroke count sits, a block's own control sits. Requested 2026-09-02 by John.

Area

study

Name

「ていけいのてほん」 · 定型の手本 Teikei no Tehon · one header, and every card wears it

The filing marks now reach search, the subject pages and the glyph viewer, but not the explorer grids themselves - JLPT, grades, level and study each draw their own cards and rows. SubjectCards has renderUnder and SubjectRows has renderAfterRow, so this is passing slots rather than changing the shared pair. Split out of add-to-lists-everywhere 2026-09-02.

Area

study

Name

「つんでおく」 · 摘んでおく Tsunde Oku · picked as you pass, and kept

Two switches on the stroke panel, both on and both remembered. Outline shipped a release earlier showing only while a single stroke was picked, on the reasoning that the full animation already draws its faint character - which read as the button disappearing when you left that view, and took the choice away where somebody might want the drawing clean. One switch now governs the faint character in both views. Replay was a one-shot button, so watching a stroke order twice meant finding it and pressing it again; it is a switch too, and the character draws itself over and over with a second's rest between runs - measured at 1,018ms in the browser. Without that rest the last stroke lands and the page blanks in the same instant, which reads as a glitch rather than a repeat. Requested 2026-09-02 by John.

Area

study

Name

「ちょっとひといき」 · ちょっと一息 Chotto Hitoiki · a short breath between runs

Studying one stroke drew that stroke and the ones before it and nothing else, so a first stroke was a single mark in an empty box with no sense of where in the character it belonged. The animation already draws the finished character faintly underneath and deliberately drops it the moment a stroke is picked, because an outline of everything is also what makes not-drawn-yet look drawn - so this is a mode rather than a change of default. Outline sits beside Replay and Numbers, appears only while a stroke is picked since the full drawing already has its outline, is off unless asked for, and is remembered per device. Requested 2026-09-02 by John.

Area

study

Name

「せおうしたじ」 · 背負う下地 Seou Shitaji · the faint ground a stroke is laid on

The stroke picker offered the numbers and nothing else, so following a character stroke by stroke meant aiming at a different target for every step - and a twenty-nine stroke character wraps its numbers into a block to search through. Previous and next sit together in front of the numbers, so one repeated press walks the whole character from a control that does not move; they were tried at either end of the numbers first, which put them half a panel apart as soon as the row wrapped. The ends join up rather than going dead, because somebody who has just reached the last stroke wants to watch it built again and a stopped button sends them back to hunt for 1. The opener and the numbers also share one row now instead of stacking. Requested 2026-09-02 by John.

Area

study

Name

「しめすいっぽ」 · 示す一歩 Shimesu Ippo · one step, then the next

A kanji page said how the character is drawn and what it means but not what it is made of, while the RADKFILE index behind the radical search already knew the parts of 6,355 characters. Written with sits under the stroke box: each part as it is written, its English name where the dictionary has one, and a link into the radical search for everything else built from it - which is the question a reader has next. Its own block with its own credit, because the data is EDRDG's rather than WaniKani's and every block on this page names its source. The index is stored radical-first, since that is the direction a search runs, so the page's question is answered by testing 253 memberships rather than by keeping a second copy the other way round. Reachable on its own at /kanji/<character>/parts, and absent for a character RADKFILE does not cover. Requested 2026-09-02 by John.

Area

study

Name

「こまかいぶひん」 · 細かい部品 Komakai Buhin · the small parts a character is made of

A list can be marked complete or studied by its owner, and carries a grade the system assigns after a study or review pass over it - how much of the list is known, and how well it went. Shown on the card and the list's page. Ties to practising from a list and the who-knows-what overlay. Requested 2026-09-01 by John.

Area

study

Name

「けいかのけじめ」 · 経過のけじめ Keika no Kejime · how it is going, and the line drawn under it

Surfaces that show items WaniKani never taught number their rows -1, -2, -3 down the page, so React has a key and a selection has something to hold. The review stats card passed that straight to /api/study/<id>/subjects/<subjectId>/history, which answered 400 twice on every open of a home-made list item - reproduced on a list of three such items before the fix and gone after it. The card now draws nothing for a row with no subject behind it, because there is no review history to show. The rule the filer already kept is one shared predicate, isCatalogSubjectId, rather than a > 0 test written wherever somebody remembered: the same mistake stopped filing from saving at all, because a single stand-in made the list route reject the whole PATCH.

Area

study

Name

「きかないかりばん」 · 聞かない仮番 Kikanai Kariban · the stand-in it stops asking about

Study mode could not be turned off by any means. The hook returned true from its initialiser and ignored what the page had resolved from the address; it answered studyMode=off and studyMode=0 with setStudyMode(true), so the off switch turned it on; it read a stored value of any kind, including the '0' it wrote itself, as on; and it rewrote the address to studyMode=on whatever the member chose. The menu had no way out either - all four entries pick a review behaviour and every one of them turns study mode on - while the button beside it had always drawn an off state nobody could reach. So every surface gated on it stayed hidden: the JLPT explorer showed an ellipsis where meanings and readings go, and never showed a kanji's compounds at all. The decision is now one pure function, resolveStudyMode, with the address beating the browser's memory beating what it already was, and the menu offers Off above the four behaviours.

Area

study

Name

「かえせるきりかえ」 · 返せる切り替え Kaeseru Kirikae · a switch that goes back

On a shared list, each member's standing on every item - not started, learning, known - as a small overlay, so a family can see who has the week's kanji down and who needs a hand. Uses each member's existing visibility setting. Suggested 2026-09-01.

Area

study

Name

「おたがいのおぼえ」 · お互いの覚え Otagai no Oboe · what each other has learned

The JLPT explorer asked for a kanji's review history under whatever the address bar held: the first segment after /users/ is the member's nickname or slug, never their account id, so every request went to /api/study/<nickname>/... and answered 404. Show review stats could never fill in, and opening any kanji logged a failed request. The account id was already a prop the component used for its other fetches, one line away from the guess. The two per-kanji lookups - this history and the compounds - now live together in useSelectedKanjiDetails, each result stored beside the kanji it belongs to, so arrowing down the list cannot show the previous kanji's history under the new kanji's glyph.

Area

study

Name

「えるあてさき」 · 得る宛先 Eru Atesaki · the address it actually reaches

The viewer is where a reader decides a character is worth keeping - they have opened it, read the meaning, seen the words it appears in - and it was the one surface that could not act on that. The saved lists now sit under its title as a rail that scrolls, with the lists the character is already on first, so the strip costs the same space whether somebody keeps two lists or forty. Trouble and favourite stay inside the glyph, where they already were. The explorer grids are add-to-lists-explorers.

Area

study

Name

「いれるいばしょ」 · 入れる居場所 Ireru Ibasho · the place to put it, from where you found it

The kanji page gained an address for each of its parts; the radical and word pages, which are built from the same blocks, kept answering only as a whole. Both now take the same section segment - /radicals/leaf/related, /vocabulary/水泡/examples - from one shared registry and one shared address module, so the three public subject pages cannot drift into three different answers to the same question. A section a subject has nothing under is a 404, and every section page names the subject's own page as canonical. Requested 2026-09-02 by John.

Area

study

Name

「あけるあいかぎ」 · 開ける合鍵 Akeru Aikagi · one spare key, and it opens all three

The kanji page shows everything about a character at one address, so there is no way to hand somebody the stroke order and nothing else. Each panel gets a path of its own under the character - /kanji/<character>/stroke shows the stroke order alone, and the other panels follow the same shape - built from one registry so a section is defined once and rendered identically whether it is on the whole page or on its own. An unknown segment is a 404 rather than a silent fall back to the whole page. The same shape then covers radicals and words, and the share control on each panel points at that panel's own address. Requested 2026-09-02 by John.

Area

study

Name

「ろうかのいっしつ」 · 廊下の一室 Rouka no Isshitsu · one room off the corridor, with a door of its own

Lists, like all data on the site, are sortable, reversible, pageable, scrollable, searchable, cached and fast. The Lists page and the list viewer get search, sort by name, size and last change, and paging once a list outgrows a screen; the same standard applies to every surface that lists data. Stated 2026-09-01 by John as a standing rule.

Area

study

Name

「れっきょ」 · 列挙 Rekkyo · listing things out, one page at a time

The list page appended the print flag to a worksheet address by hand, which broke the moment a shared list's address started carrying its key: the two queries ran together and Print 404ed on exactly the lists worth sharing. The card had the inverse hack - stripping a trailing flag to derive the worksheet link - so when the prop stopped carrying one, both its pills became the same link and Print silently stopped printing. listPrintHref and printNowHref are the only things that spell it now, with a test that fails if anything else does. Found 2026-09-02.

Area

study

Name

「るいじのるいすい」 · 類似の類推 Ruiji no Ruisui · inferring one link from another because they look alike

The stroke order panel plays the whole character and nothing else: there is no way to hold a single stroke still and look at it, which is what somebody copying the character actually needs. A row of numbers opens under the header - closed by default, so the panel stays as quiet as it is now - and picking a number draws that stroke alone: the strokes before it in grey, the strokes after it blank, the chosen one in colour and animated so the pen direction is visible. Requested 2026-09-02 by John.

Area

study

Name

「りんしょのいっかく」 · 臨書の一画 Rinsho no Ikkaku · one stroke of the model you are copying

A worksheet from a list you do not own. The sheet is built at your own address, so a list belonging to somebody else is named in the path - /practice/list/mika/week-1 - rather than looked up on your own shelf, where the same name means a different list. Reachability is the list's own: public to anyone, unlisted to whoever holds the key, private to nobody. Games from a shared list are list-game-from-shared.

Area

study

Name

「らいほう」 · 来訪 Raihou · a visit, calling on somebody else

Strokes, Grades and the WaniKani library each draw their own card grid, and side by side they read as three different products: different card shape, different chrome, different placement of level and status. There is already a shared pair for lists of subjects (SubjectCards and SubjectRows) and a UnifiedExplorerCard, but these three do not all go through them. Put the general look in one component and let each page pass what is genuinely its own - a stroke count, a JLPT pill, an SRS stage - rather than restating the whole card.

Area

study

Name

「よそおうひとくみ」 · 装う一組 Yosoou Hitokumi · one set, dressed alike

GradeKanjiBoard seeds two pieces of state from localStorage inside the useState initializer, so the server renders the default and the browser flips it after hydration. The reveal mode is the visible one: every reading paints and then hides. Strokes shares the view-mode half of the pattern and does not flash, which points at the reveal mode as the cause. No page under Learn should flash.

Area

study

Name

「やむちらつき」 · 止むちらつき Yamu Chiratsuki · the flicker that stops

A copy is a snapshot, and the person it was taken from keeps adding to theirs. The two ways forward were both wrong: follow the original and you no longer own what you are looking at, copy it again and you lose everything you changed. A copy remembers its source, so it can simply ask - and the answer sits at the top of the copy's own page: what has been added since, as characters rather than a count. Item by item, which is the point: somebody who copied a class list and dropped half of it does not want the half back, so every item is a button that takes only itself, with Take all there for the common case of agreeing with everything. Taken items are appended in the copy's own order rather than merged into the source's, because the member arranged their list; anything already held is left exactly as it is, notes included. Removals are never carried across - a friend tidying their list is not grounds for deleting from somebody else's. A source that has gone private since answers as unreachable rather than as empty, because "nothing new" and "you can no longer see this" are different things to a member deciding whether to keep a copy. The panel draws nothing at all when there is nothing new, which is almost always: a list that says "0 new" every time it is opened is a list that has learned to be ignored. Suggested 2026-09-01.

Area

study

Name

「もちかえるついか」 · 持ち帰る追加 Mochikaeru Tsuika · the additions you take home from the original

Printing a list gave you squares to trace and nothing else. That is half of what somebody printing a list wants: the other half is the list itself - the characters with their meanings and readings - to keep beside the desk or hand to somebody. Reference is a third sheet beside Trace and Stroke order, at the same address with a mode, so a chosen sheet is still a link. It is a table rather than a grid, its readings are on by default because they are the point of it, it credits KANJIDIC2 rather than the stroke art it does not draw, and it does not tell a phone to find a bigger screen - reading is the one thing a phone does well. The columns collapse to one at 393px instead of setting each reading a character per line. The tracing half of this shipped as 0.291.0; this is the rest of it.

Area

study

Name

「むすぶしおり」 · 結ぶ栞 Musubu Shiori · the list, tied to what it means

The grades search asked a learner to already know the character they came to look up: every other search here suggests what its surface holds, and this one suggested nothing. It offers the grade's own kanji with their meanings now, as a plain suggestion list on the box that was already there. All of them, in the order the grade teaches them - the search reads the whole grade, so a partial list would be a half-truth about what typing will find, and junior high's 1,110 characters are exactly the grade where knowing what is in there is worth the most. Reported 2026-09-02 by John.

Area

study

Name

「ほしいめいぼ」 · 欲しい名簿 Hoshii Meibo · the list you wanted to see

Searching 5,000 offered the kanji for five and for zero, so a price put 零 at the top of the results. The reader matched runs of digits, and a thousands comma broke 5,000 into 5 and 000 - each read as its own number. The money parser had always dropped that comma before reading an amount; the same rule is here now and the two agree about what a number is. A comma between other digits is somebody listing two numbers and is left alone, since joining those would invent a third. The other half is why the query looked broken even spelled correctly: 五千 is the right answer to 5000 and no catalogue holds it - the spelling of a compound number is not a subject anywhere - so searching it exactly found nothing at all. The characters it is written with are offered alongside the spelling, and they are all taught: 5000 answers with 千 and 五, and 24 with 十, 二 and 四. A number written with a single character still answers with that one character only. Reported 2026-09-02 by John.

Area

study

Name

「へんなてん」 · 変な点 Hen na Ten · the dot that was reading as a break

A list could be turned into a writing sheet only from its card on the shelf, and only as a picked sheet carrying every one of its characters in the query string - which broke on a long list, went stale the moment the list changed, and could not be sent to anybody as a link to “my Week 1 sheet”. The page you actually read a list on offered nothing at all. A saved list is a practice source of its own now, addressed by name at /practice/list/week-1, and the sheet is titled with the list's name so a printout says which list it is. Both surfaces - the card and the list's own page - offer Worksheet and Print, the second opening the print dialog on arrival; Burned offers neither, since tracing what you already know is not practice. The list page's search box takes a line of its own on a phone, where five controls beside it had squeezed it down to four characters. Requested 2026-09-02 by John.

Area

study

Name

「ひとふではんし」 · 一筆半紙 Hitofude Hanshi · a single brushstroke on writing paper

Two reported together. Escape emptied the search box and then did nothing: the panel left open after the clear is the recent items rather than the suggestions, and only the suggestions were in the condition, so the second press had nothing to act on. Both count now - clear, then close. And the stroke-order sizes were growing the wrong thing. The numbers are drawn inside the character's own coordinates, so asking for a bigger drawing made every number bigger by the same amount and they collided exactly as before, which is the opposite of what enlarging it was for. They are sized against the drawn width instead, so they come out the same height on screen at every setting and the extra room goes where it was wanted - between them. The largest size grew from 320 to 370 pixels and drops the four printed faces beside the drawing: somebody who has asked for the character as big as it goes is looking at the strokes, and four printed copies of it are what that width was being spent on. Reported 2026-09-02 by John.

Area

study

Name

「はなれるあいだ」 · 離れる間 Hanareru Aida · the space that opens between the numbers

Every card in a list carried "Add a note" printed under it in plain text and an × sitting on it, so a grid of characters read as a grid of controls and the thing the page exists to show came third. Both wait for the pointer now: they appear on the card under the cursor and on whatever has keyboard focus, and a note that has been written stays visible always, because that is content rather than a control. The × was drawn with the letter x, which read as part of the card; it is a proper mark in a circle that turns red where it will act. Reported 2026-09-02 by John.

Area

study

Name

「のぞくぼたん」 · 覗くボタン Nozoku Botan · the button that waits until you look

The stroke animation is drawn at one fixed size, and at that size the stroke numbers are too small to read - the whole point of turning them on. It takes a small S/M/L control, remembered per device. Three things beside it are out of true: the drawing is not centred in its row because the printed faces on one side and the buttons on the other are different widths, the Replay and Numbers labels sit left in buttons stretched to a common width, and the dictionary fact pills do not centre their text. Reported 2026-09-02 by John with screenshots.

Area

study

Name

「ねばりのばんごう」 · 粘りの番号 Nebari no Bangou · the numbers you can finally stay with

Four things about the search row, reported together. Find by radicals and Add to lists were two rows of options stacked above the results; they are one row now, and the radical picker takes that row over while it is running rather than opening a second one - everything that changes how the search behaves is in the same place. The dropdown is as wide as the field it hangs from, since a fixed width left it narrower than the open box, which looked like a second control and wrapped the options onto two lines. Opening the header's box made the section links wrap to a second line while squeezing the field; the links now step aside entirely and the box takes the whole row, because somebody who has started typing is not reaching for Learn and the links are one Escape away. Escape empties the box before it closes it - one key, two things to undo, and closing first left the box still full so the next focus reopened the search somebody was trying to leave. And the radical grid can be resized: smaller, larger and reset, remembered per browser, starting larger than it shipped at. A grid of 253 characters is read by shape - 冫 and 氵 differ by one stroke - so how big it is drawn decides whether it can be read at all, and that answer is different on a phone, at a desk and for older eyes. Reported 2026-09-02 by John.

Area

study

Name

「ぬけてひろがる」 · 抜けて広がる Nukete Hirogaru · the links step out and the box widens

The radical command could only be written in Japanese, which left the typed half of the feature useless to exactly the reader it was built for: somebody looking up a character they cannot read is usually in the same position with its parts - they can see the shape and have no way to type it. :radicals sun + moon now answers with the same thirty-one characters as 日 + 月. Names come from two sources, best first, because neither covers the set alone: the kanji dictionary for the 247 of 253 radicals that are also kanji, which is the accurate answer and costs nothing since the file is already shipped; and WaniKani's own radical names for the 199 whose character is in this set, because a member taught "fins" for ハ will type what they were taught. WaniKani's are teaching mnemonics rather than classical names - it calls 久 a raptor cage - so they answer second and never over the dictionary. A name that reaches neither answers with nothing rather than a neighbour: WaniKani draws some of its radicals instead of writing them, so "beggar" names a picture and no character here. The command's own reading changed with it. A space is inside "long time" and between 日 and 月, so it is no longer a separator: parts split on a plus or a comma, and then a part of Japanese script becomes one radical per character while a part of Latin letters stays one name however many words it holds. Requested 2026-09-02 by John.

Area

study

Name

「にほんごいがい」 · 日本語以外 Nihongo Igai · a radical named in English too

Fifteen of WaniKani's 504 radicals arrive with no character, so every surface fell through to the slug and a review of 旅 asked the member to recognise the English word "tofu". WaniKani draws those as artwork on a gated CDN that is its copyright either way. Nine turn out to be ordinary Unicode characters WaniKani never filled in, found by subtracting a kanji's known components from WaniKani's own decomposition and checking the remainder against two independent databases; the other six are shapes Unicode does not encode at all and take private-use codepoints. All fifteen are carved out of Noto Sans JP rather than drawn - Hills is 之 without its top stroke, Cactus the bottom of 虚 - so they match the site's own Japanese face because they are that face. The font ships at two weights and is scoped by unicode-range, so it is consulted for those fifteen codepoints and never disturbs the gothic, mincho, textbook or brush switch.

Area

study

Name

「なぞるかたち」 · なぞる形 Nazoru Katachi · a shape traced for every radical

The header's search was three sizes smaller than the results page's, which made one control read as two different ones - and it had grown a command language that needs room. Both boxes are the same height and the same type now, the header's collapses narrower than before when nothing is in it, and it opens far wider when it is being used, wider again while a command is in it. On a narrow desktop the section links step aside while the box is open, since somebody who has started typing is not reaching for Learn. And on the search page the header's box is gone entirely - it was a second field asking the same question a hand's width above the first - with Search named in the row instead, where the section a member is looking at belongs. Reported 2026-09-02 by John.

Area

study

Name

「とういつのおおきさ」 · 統一の大きさ Touitsu no Ookisa · one size for both search boxes

A member could file a kanji from a search row but not from the kanji's own page, which is where you decide a character is worth keeping. The public kanji, word and radical pages carry a strip under the card with the same marks the search column uses - trouble, favourite, and a chip for each saved list - and a visitor gets the offer to join instead. Part of add-to-lists-everywhere, which keeps the glyph viewer and the explorers. Reported 2026-09-02 by John with screenshots.

Area

study

Name

「てもとてちょう」 · 手元手帳 Temoto Techou · the notebook kept at hand

Three things reported on the radical command. The grid gave every stroke count a row of its own, which wasted almost all of the width on the counts that hold three radicals - fourteen strokes is 鼻 and 齊 - and made the picker twice as tall as it needed to be, which is the space the answers wanted. The radicals now flow left to right in one run with the stroke count as a marker between groups, the way paper dictionaries print it, and the grid is ten rows rather than fourteen. The command was also the only way in, and nobody guesses :rad, so there is a row under the search box now - Find by Radicals, or type :rad - and the button writes the command rather than opening a panel of its own, so pressing it and typing the words land in the same place. It sits in the layout under the input rather than inside the dropdown, because the results page keeps its dropdown shut on focus so a seeded query cannot cover the results it just produced, which would have left the row permanently invisible there. And a bug: after a radical search, focusing the box again showed nothing until a key was pressed. A command in the box is the picker's own state, so focus now brings both the picker and its answer back. Reported 2026-09-02 by John.

Area

study

Name

「つめてならぶ」 · 詰めて並ぶ Tsumete Narabu · the grid packed into fewer rows

The radical lookup shipped as a pill inside the search field that opened a dialog over the page, and the two halves never agreed: the query was in one place and the answer in another, a picked radical left no trace in the box, and nothing about it could be typed, copied or linked. It is a command now. Typing :rad, :radical or :radicals opens the grid under the search bar rather than over the page, picking writes :radicals 日 + 月 into the box, and the matching kanji come back as ordinary search results underneath - with the kind tabs, the keyboard and the filing column every other search already has. The query is the whole state, so it can be typed by hand, pasted to somebody and walked with the back button. The pill is gone from the field and the search bar grows while a command is in it. The reason a dialog had been chosen in the first place turned out to be a bug of its own, and a much older one: the search page's shell was clipped with overflow-hidden, so every dropdown was cut off at the shell's height. A search for "water" found ninety-one results, showed eleven and drew three. Clipping sideways is what that was for, and it now clips sideways only. Reported 2026-09-02 by John with screenshots.

Area

study

Name

「ちゅうもんのけんさく」 · 注文の検索 Chuumon no Kensaku · the search you type as a command

A list of glyphs says what to study and never says why. The mnemonic that finally made one stick, the sentence it was met in, the mistake it keeps causing - those are the reason a member's own list beats a generated one, and they had nowhere to live. Every item on a saved list can now carry a short note, written from the list itself while looking at the thing it is about, and shown to every reader of the list in both densities. The invitation to write one appears only while editing, so a list nobody wants to annotate stays a list of glyphs. One item at a time rather than a save over the whole list, because a whole-list write would race with any other change the member has open. The correctness half is invisible and was the larger part: saving a list deletes every row and writes them again, so a caller that read only the columns it cared about would have wiped the notes on every other item the moment somebody removed one. Every caller that rewrites a list now reads the note first, and a test fails the build if a new one forgets - and while that was open, an item's authorship stopped being reassigned to whoever last touched the list. Suggested 2026-09-01.

Area

study

Name

「たったひとこと」 · たった一言 Tatta Hitokoto · one line saying why it is here

The way to look up a character you cannot read: you cannot type it and you do not know its readings, but you can see 日 on the left and 月 on the right. Every search box now carries a Radicals button - the header, the phone sheet and the results page, since they are one field - and it opens the classical grid: 253 radicals in stroke-count order, the ones paper dictionaries index by. Picking narrows on all of them rather than any, so 日 with 月 is the thirty-one characters holding both, commonest first, each one a link to its page. Radicals that appear in none of what is left are dimmed rather than left as routes to an empty list, which is the whole ergonomics of the thing. The answer is pinned above the grid so a pick made at the bottom of it is still answered where you can see it. The radicals are RADKFILE's - Michael Raine's analysis of the JIS kanji, kept by the EDRDG under the same share-alike terms as the dictionary - covering 6,355 characters, rather than WaniKani's teaching radicals, which are named for what they look like and only reach the two thousand kanji it teaches. It is a dialog rather than a dropdown because 253 buttons and their answer do not fit under a field: hung from the header on a phone it ran six hundred pixels past the bottom of a screen that clips. Requested 2026-09-02 by John.

Area

study

Name

「そなえるぶしゅ」 · 備える部首 Sonaeru Bushu · the radicals a character is made ready from

The line of facts shipped an hour earlier was drawn as a list, and two of the three surfaces that show it put it inside a paragraph - the list page header and the followed-list card. No browser allows a list inside a paragraph, so the markup the server sent and the markup the browser built disagreed, React gave up matching them and rebuilt the whole tree on hydration. It reads identically and is now spans, which are legal wherever prose is. Found by reading the browser console rather than the page, which is the only place this kind of fault says anything at all.

Area

study

Name

「せいれつなおし」 · 整列直し Seiretsu Naoshi · the facts line up where they belong

A list card said one thing about itself - when it last changed - so a reader deciding whether a shared list was worth opening had nothing to go on: not how big it is, not whether anybody else keeps it, not whether it is still being added to. Every list now carries the same facts wherever it is described: when it was made, when it last changed, how many people follow it, and how many times it has been copied and shared. They are drawn by one component, so a list cannot describe itself one way on its card and another on its page - the list page had this vocabulary already and the cards had none of it, and now there is one. A fact that has not happened is left out rather than printed as a zero, because "copied 0 times" is a sentence about failure and a list nobody has copied yet simply does not mention copying. The follower count is new data: the lists query counts subscriptions alongside the rows. Two details that only showed up on screen: the separator trails its fact rather than leading the next one, or a wrapped line starts with a dot nobody meant, and the card no longer prints its size twice. Requested 2026-09-01 by John.

Area

study

Name

「すぐわかるゆらい」 · すぐ分かる由来 Sugu Wakaru Yurai · where a list came from, at a glance

Japan counts in ten-thousands, and every price a learner meets is written that way: rent is 8万円, a salary 400万円, a flat 3,000万円. The search read the digits and dropped the unit, so 400万円 came back as four hundred yen - wrong by four decimal places on exactly the figures somebody moving to Japan needs to understand. 千, 万, 億 and 兆 are read now, and they stack the way they are written, so 3億5000万円 is three hundred and fifty million and 5万3千円 is fifty-three thousand. Units that climb rather than fall are a typo and answer with nothing, because a confident wrong figure is worse than none. The Latin shorthand is the same idea in the reader's own writing - 20k yen, 1.5m yen, $50k - and a magnitude letter is only one where a word does not follow, or the M of 20 MXN would read as a million and leave XN behind as the currency. The size cap moved with it: it was a trillion, which answered Japan's national debt of 1,300兆円 with nothing. A magnitude still needs a currency beside it, so 20k and 5万 on their own remain numbers rather than prices. Requested 2026-09-02 by John.

Area

study

Name

「しききんのけた」 · 敷金の桁 Shikikin no Keta · the deposit read at the right number of zeros

500¥ and 100 CAD converted while 500 yen and C$100 answered with nothing, which is the wrong half: the code is what a bank uses and the sign is a keystroke most keyboards hide, so the words are what a reader actually types. A currency written as a word is read now - yen, euros, pounds, francs, won, rupees, pesos and the rest, singular or plural, since the lookup retries without a trailing s - and so are the letters in front of a dollar sign, which is how a Canadian price is written wherever both dollars are in the room: C$100, US$20, A$50, HK$300, R$50. A word can be two currencies exactly as a sign can, so 100 dollars and 20 bucks answer in Canadian and American together, while a peso falls back to the Mexican one and a krona to the Swedish, each being the one a North American reader is likelier to have met. A number beside an ordinary word is still not a price: chapter 3, level 5 and Heisei 3 name no currency and answer with nothing, which is what leaves the era converter its own query. Reported 2026-09-02 by John.

Area

study

Name

「このみのかきかた」 · 好みの書き方 Konomi no Kakikata · a price written the way you write it

Typing "Heisei 3" into the header box answered with ten rows about the number three and nothing else. The era and money answers had been built and shown only on the full results page, while the dropdown under the box - which is where a query is actually typed - called an endpoint that never asked for them. The answer now rides back with the rows rather than taking a request of its own, since it answers the same query, and sits at the top of the dropdown on the desktop and in the phone sheet alike: the value, the Japanese form, and the question it answers. It shows even when no catalogue matched, because "Taisho 5" is 1916 and no row exists for it - a dropdown that said no results to a question it had just answered would be hiding the answer behind its own emptiness. No rate history there and no attribution line: those are a table and a sentence, and the dropdown is ten rows under a search box, so it asks for today's rate only and one cached request does not become six on a keystroke. Reported 2026-09-02 by John.

Area

study

Name

「くわえるいちぎょう」 · 加える一行 Kuwaeru Ichigyou · one line added above the rows

23 EUR and 100 USD converted while $14.40 CAD and $20 matched nothing at all, so the one form the feature was asked for was the one it missed. The dollar sign now names the dollar family: a code beside it picks from that family, so $14.40 CAD is Canadian and $1,234 AUD is Australian; a code outside it contradicts the sign and names nothing, so $1,500 JPY answers with nothing rather than guessing past the contradiction; and on its own the sign falls back to the two currencies this site is written for and answers in both, which is the same choice a yen amount already makes coming the other way. Reported 2026-09-02 by John.

Area

study

Name

「きまるどる」 · 決まるドル Kimaru Doru · which dollar you meant, settled

Searching an English meaning offered the word and never the character it names. "Morning" answered with the word 朝, the radical ₙd and eight rows besides, but not the kanji 朝 - and not from any of the three catalogues that teach it. The suggestion list keeps one row per glyph, which is right when the same kanji arrives from WaniKani, the JLPT table and the school grades: a ten-row list cannot spend three rows on 日. But a glyph is not a subject. 朝 is a kanji and a word written the same way, so the word claimed the glyph and all three kanji rows behind it were read as copies and dropped; which of the two led was settled by nothing more than 3044 sorting ahead of 695 as text. Every single-character word hid its kanji the same way - 水, 山, 人, 川. The row is a subject now rather than a spelling: the same kanji from three catalogues is still one row, while a kanji and a word that share a spelling are two, because they are two things with two pages behind them. Reported 2026-09-02 by John.

Area

study

Name

「わすれないかたほう」 · 忘れない片方 Wasurenai Katahou · a word and its kanji, both kept

The site groups kanji by what a learner is doing - the WaniKani level they are on, the year a Japanese child is taught them, the JLPT that will test them - and none of those answer the question somebody has with a pen in their hand: what else is written in this many strokes? The dictionary has known every character's stroke count since the day it was ingested, so /strokes lists all thirty-three counts with how many kanji each holds, and /strokes/12 is those nine hundred and twenty-five characters, commonest first. A Common only filter keeps the two hundred and fifty-seven a newspaper actually uses, which is the difference between a page to study and a page to scroll, and long counts page. The count is a path so it can be linked and reloaded; what is on the page - the filter, the page number - stays in the query. A signed-in member gets the same add-to-lists column search has, so a character met here goes straight onto one of their lists. It sits under Learn beside the maps. Requested 2026-09-01 by John.

Area

study

Name

「ろくせんかぞえ」 · 六線数え Rokusen Kazoe · counted by the strokes it takes

The pills that stand for a kanji or a word turn up in several places - the words a kanji is used in, the related groups on a subject page, a paste being turned into a list, a list being edited - and they had drifted: some carried the reading and the meaning, some carried the glyph alone. They are one component now, and every one of them can say what it is: the reading and the meaning under the glyph, with one control that shows or hides that text everywhere at once, remembered per browser and on by default. The paste preview gained the words it was missing, so a chapter comes back as items a member can read rather than a wall of characters to guess at. The pill also decides for itself what is Japanese, so a radical WaniKani draws by name is no longer set in a Japanese face. Requested 2026-09-02 by John.

Area

study

Name

「やさしくよめる」 · 優しく読める Yasashiku Yomeru · every pill says what it is

Opening the JLPT explorer as an account with no WaniKani connection answered We could not load this user drilldown. The page asked WaniKani for the member's own kanji with an empty token, took a 401 and threw, so the one page that is a table of every JLPT kanji - the part of the site that needs WaniKani least - was the one page a member without WaniKani could not open. It asks only when there is a connection to ask with, and a token that has since been revoked is treated the same way: the explorer opens, and the member's own progress is simply the part that is not there. Reported 2026-09-02 by John.

Area

study

Name

「むかうさき」 · 向かう先 Mukau Saki · somewhere to go, with or without a connection

A list opened as a panel from one control and as a page from another, so which view a member got was a mystery, and the two had drifted apart. There is one view now, and it is the page. Trouble, Favourites and Burned have addresses of their own - /users/<member>/lists/trouble and its two siblings - and every card, name and button leads to a page. The panel is gone, and nothing went with it: the page draws the same subject cards and rows every other surface uses, with the member's own SRS state and tags on them, and it carries choosing and the practice sheet, the glyph viewer, an Edit toggle that takes an item out - untagging a built-in list, dropping a row from a saved one - and the WaniKani offer on Burned. Opening somebody else's list shows how far you have got with what is on it, since the state read is the reader's own. Reported 2026-09-02 by John.

Area

study

Name

「みっつひとつ」 · 三つ一つ Mittsu Hitotsu · three ways in, one view

The auto lists sat under Explore as Kanji lists while a member's own lists sat under Lists, on a page that looked nothing like theirs. Every list lives under Lists now: the built-in two, the sets you saved, the lists you follow, the auto lists in full - all of them on one page, in sections, with a star on the auto lists you follow. Explore has lost its Kanji lists entry, and Lists leads a visitor with no page of their own to the auto lists, which are the part of Lists that belongs to everybody. The three list surfaces also shared out one search field, so the viewer, a list's page and an auto list's page all clear with an X and offer the list's own items as you type, rather than one having both and two having neither. They are called Auto lists now, which is what John called them. Requested 2026-09-02 by John with screenshots.

Area

study

Name

「まなぶあつまり」 · 学ぶ集まり Manabu Atsumari · everything to learn from, gathered in one place

Two or more of your lists made into one, from the Lists page. They are numbered as they are picked, because the order is the answer - merging Week 1 and Week 2 should read as Week 1 then Week 2 - and the panel says how big the result will be and how many items the lists shared before anything is written. Clearing the sources away afterwards is a choice rather than the default, and goes through the same rule the delete button uses, so a source somebody else follows is archived rather than deleted out from under them. Requested 2026-09-02 by John.

Area

study

Name

「ほんのいっさつ」 · ほんの一冊 Honno Issatsu · several lists, just the one volume

Paste a passage, a handout or a page of a book into the Lists page and every word the catalogue knows and every kanji in the text comes out, words matched longest-first so a run reads as the words it holds rather than a string of characters. What comes out is a preview to massage before anything is written: each item is a button that takes itself out, a whole kind can be dropped at once, and the list is saved only when it is named. Pasted text is treated as hostile - cut to a cap of twenty thousand characters, stripped of control characters, never stored, never rendered as anything but the words in the box, and answered only with items the catalogue already knew, so markup or SQL in a paste comes back as nothing at all. Requested 2026-09-01 by John, restated 2026-09-02 with the safety and massage requirements.

Area

study

Name

「へらしてつくる」 · 減らして作る Herashite Tsukuru · cut down first, then made

Seventy-two lists nobody owns, gathered at /lists: every JLPT level N5 to N1, every Japanese school grade and the junior-high set, and every one of WaniKani's sixty levels. They are not snapshots but live - a question stored as a list rather than rows, asked again each time one is opened - so Grade 1 is whatever the grade table says today and a WaniKani level is what it teaches now, radicals then kanji then the words that use them. Each has a public page of its own with the same rows, kind chips, search and burned filter a member's list gets, and the two ways to keep it: follow it and it stays current under Lists that keep themselves on your own page, or copy it and it becomes rows of your own, to cut down to half of Grade 1 or to Grade 1 without what you already know. The index sits under Explore as Kanji lists. Requested 2026-09-01 by John.

Area

study

Name

「ふえるもくろく」 · 増える目録 Fueru Mokuroku · a catalogue that keeps itself

Three things reported together. The search page had become unreadable: every row carried a kind pill, a level pill, a Strokes chip and - because the filing column had been opened once and was remembered for ever - two tag marks and a chip per list. Filing is now held for the sitting rather than the browser, so it is off unless it was opened in this tab, and the Strokes chip is gone, since the kanji page a row already leads to opens the same animation in the same number of clicks. The Burned list's offer to apply WaniKani's burned items counted every burned item rather than the ones not applied yet, so it still said Apply 1,728 after applying them; it counts what is left and disappears at zero. Removing an item from a list showed a mark on every card with no mode to explain it, and is now behind an Edit toggle in the viewer. And searching inside a list read the raw string, so water found 水 and mizu found nothing while the search box on the same page found both: both list surfaces now read a query the way search does, romaji turned to kana and tried against the readings. Reported 2026-09-02 by John with screenshots.

Area

study

Name

「ひろえるひびき」 · 拾える響き Hiroeru Hibiki · a sound you can pick up, however you type it

Three things John found opening his own lists. Opening a saved list offered Choose and no way to take an item out - the remove button the Trouble and Favourites lists always had now works on a saved list too, in place. Choosing was remembered across lists, so closing one list and opening another opened it still choosing, with the first list's picks counted; choosing resets when a different list opens. And the viewer's title was the bare name with a '10 items' pill in the toolbar, so nothing said this was a saved list: the title now says Saved list above the name and carries the count beside it, and the search box has a clear button and offers the list's own items as they are typed, like the other search boxes. Reported 2026-09-02 by John.

Area

study

Name

「はずしてもどる」 · 外して戻る Hazushite Modoru · taken out, and back where you started

A third built-in list beside Trouble and Favourites: Burned, for the items you know so well you never need to read them again. It starts empty for everyone, and is marked by hand with a third mark - a small flame - on every glyph the other two marks are on. A member with WaniKani connected opens the Burned list and sees an offer with the count on it, Apply 157 burned from WaniKani, applied on the click and again whenever they like, never silently on sync. Applied, a Hide burned toggle on the list viewer and on every list page takes those items out and says how many went, so a list never looks shorter than it is without saying why. There is no practice sheet for what you know. Requested 2026-09-01 by John.

Area

study

Name

「のけるきおく」 · 除ける記憶 Nokeru Kioku · what you already know, set aside

A private list nobody else has seen is deleted, as it always was. A list that others hold - shared by link or public, followed, copied, or with suggestions waiting - is archived instead: closed to change for everyone, the owner included, still readable, copyable and followable by everyone who had it, and listed under Archived on the owner's Lists page with Restore and Delete for good. The confirmation says which of the two will happen before the click, and deleting for good says who loses the link. A follower whose list has gone private or away sees so on their own page. Requested 2026-09-01 by John.

Area

study

Name

「ねかせてまもる」 · 寝かせて守る Nekasete Mamoru · put to rest, and kept for those who hold it

A shared list can be opened so anyone who can see it adds to it at once, or kept locked, from a control beside the visibility. On a locked list a member's addition becomes a suggestion, typed into a box on the list's page with a note on why; taking an item out is always a suggestion, whoever asks, since a list open to additions is not open to being emptied. The owner sees the suggestions on the list itself - the item, the verb, who asked, when and why - and approves or declines each one; approving changes the list on the spot, and a suggestion that would decide nothing is never made. Requested 2026-09-01 by John.

Area

study

Name

「ぬしがきめる」 · 主が決める Nushi ga Kimeru · the owner decides

Reading somebody's list now offers two ways to keep it, for two different promises. Copy makes a list of your own from its items - named so it fits on your shelf, Week 1 (copy) if you have a Week 1 - and remembers where it came from, while the source counts one more copy. Follow keeps the owner's list as theirs: read-only, always current, listed under Following on your Lists page with the owner's name, and dropped from there. A list made private since you followed it says so. The add-to-lists column from search sits on the list page as well, so one row of a friend's list can be tagged or filed into a list of your own. Both need exactly the permission that opening the list needs. Requested 2026-09-01 by John.

Area

study

Name

「にたもの」 · 似た物 Nita Mono · a copy that is yours, or theirs to keep

A list is a first-class thing, so it has an address of its own: /users/<member>/lists/<name>, the name in kebab case, a Japanese name kept as itself. The owner chooses Private, Link only or Public from the page. Link only makes an unguessable key the moment it is chosen, and the copy button hands out the link; Public needs no key. Whoever may open the list gets the same page - the name, who made it, when it was made and last changed, how often it has been shared, kind chips, a search, cards or rows - and somebody not signed in gets one quiet card inviting them to sign in and keep it. A private list reads as absent to everyone else. Names are displayable - letters in any script, digits and a title's punctuation, stray marks dropped - and two names that would share an address are refused. Requested 2026-09-01 by John.

Area

study

Name

「なづけたじゅうしょ」 · 名付けた住所 Nazuketa Juusho · an address with a name on it

A list was a string of characters, which is why it could only hold kanji and why a word added from search became its kanji. A list holds items now, each with its kind - radical, kanji, word, sentence - named the way its own catalogue names it, with WaniKani's id beside it where there is one, so a kanji WaniKani does not teach is still an item. A card previews words whole and shows kind chips; the viewer filters by kind; the editor takes kanji and words typed or pasted, reading a run of kanji one by one and anything with kana as a word; the search filer files a word as a word. The Lists page searches, sorts by last change, name or size, and reverses. Production's two lists were copied into items with a snapshot first, item for item. The base for every sharing feature; the first of the lists programme.

Area

study

Name

「とりよせたしな」 · 取り寄せた品 Toriyoseta Shina · the items, gathered as what they are

Pressing Add to lists opened the filing column in the rows, but the dropdown kept its width, so on a desktop the chips were crushed against the meaning instead of sitting in a column on the right. The box that widens and the list that shows the column each kept their own copy of the open flag, and nothing told the box. One flag now, held once, with everything that shows it subscribed - flip it anywhere and the box and the rows change together. The cell sits inline only from the width where the box is truly wide; below that it drops to a line under the row, as on a phone. Reported 2026-09-01 by John.

Area

study

Name

「てざわりのはば」 · 手触りの幅 Tezawari no Haba · wide enough to reach the column

Choosing N5 on the JLPT explorer changed nothing about the URL, so a filtered view could not be linked, bookmarked or reloaded. The level and study explorers had already put their filters in the address when they became real routes; this one had not followed. Its levels, WaniKani filter, level and grade filters and the sticky toggle are now written to the address as they change and read back on arrival, where the address wins over what the browser remembers - so a link to N5 opens on N5 whatever the last visit chose. A filter at its default says nothing, so a plain link stays plain.

Area

study

Name

「ちょくせつたどる」 · 直接辿る Chokusetsu Tadoru · reached directly, by its address

Searching ten kanji and keeping each one meant, for every kanji: open it, find the tag on its page, go back, search the next. The results dropdown and the search page now carry an Add to lists control at the edge of the list. Opened, every row grows a quiet column: the trouble and favourite marks the explorers use, then one small chip per saved list, lit when the row is already on it and toggled with a tap. The dropdown widens to make room; in the narrower search columns and on a phone the column drops to a line under the row. Kanji the JLPT and school-grade catalogues answer with borrow their WaniKani twin's id so they can be tagged too, and a word contributes its kanji to a list rather than its kana. Whether the column is open is remembered per browser. Requested 2026-09-01 by John.

Area

study

Name

「たばねたてがかり」 · 束ねた手掛かり Tabaneta Tegakari · the clues, bundled as they are found

The stroke-order panel showed the character in two faces, Gothic and Mincho, each captioned in capitals underneath, and the captions drew the eye from the shapes. They are gone; the face's name is a hover title now. Two more faces join them, the two a learner actually meets: the textbook face (kyoukashotai), the handwriting standard every Japanese school teaches from and the one to copy where a printed form and the written form part ways, and a brush face the printed forms descend from. Neither ships with any system, so both come from Google Fonts at runtime, swapped in when they arrive. The four sit in a two-by-two grid beside the drawing, printed faces above and written faces beneath, and the stroke count pill has moved to the right of the header. Requested 2026-09-01.

Area

study

Name

「そなえたしょたい」 · 備えた書体 Sonaeta Shotai · the faces a learner meets, provided

A word page listed the kanji it was written with and stopped - a dead end with two chips on it. It now goes on to the other words built from those same kanji, easiest first, so a page about one word is a way into its surroundings. Two sources could have supplied the list: each kanji's JLPT word examples cover more characters, but an example is only a string, with no level and no page unless the catalogue happens to hold it; each kanji's WaniKani words are catalogue rows with a level and an address, and every one of them has a page. The neighbourhood is drawn from the WaniKani words, because a list of places to go beats a longer list of names.

Area

study

Name

「せつぞくのきんじょ」 · 接続の近所 Setsuzoku no Kinjo · the neighbourhood, joined up

The reading labels said On readings and Kun readings in English and nothing else, so a student asked about on'yomi on a test had never met the word here. Every label now carries the Japanese beside it - 音読み, 訓読み, 名乗り - with the romaji in a hover title. The readings under each are written the way a dictionary writes them, on in katakana and kun in hiragana, whichever way the source stored them: KANJIDIC had it right, the school tables sometimes did not, WaniKani never did, and one conversion now settles it for every page. Where there is room the romaji sits beside each reading in a lighter weight, for the reader still weak in kana, and steps aside on a phone. One component draws the line - the dictionary block, the grade cards, the grade list's two lanes, the JLPT detail and the stroke-order header all take it - so the label cannot be spelled two ways and the script cannot differ between pages.

Area

study

Name

「きこえるきごう」 · 聞こえる記号 Kikoeru Kigou · a reading you can hear as well as see

Every section of a subject page names where its content came from, and each named it a different way: the stroke-order panel with a rule across the full width and a small centred line beneath it, the dictionary facts with a left-aligned line at another size, the compounds and the relations and the mnemonics each at their own indent. Five cards down a page, five feet. The stroke-order foot was the right one and is now the only one. The credit component draws it, the block shell places it at the foot of the card with the rule running edge to edge, and every section that shows borrowed content sits on that shell - sentences, dictionary facts, the word and radical header, the stroke-order panel itself, which had been drawing its own copy of the style it inspired. The stroke animation alone keeps an inline line, because it draws inside another surface and has no edge of its own to put a rule on.

Area

study

Name

「えらんだあしもと」 · 選んだ足元 Eranda Ashimoto · one footing, chosen for every card

Every search result lands on a kanji's own page, and that page knew less about the character than the study viewer behind the sign-in wall. It had the strokes, the dictionary facts and the sentences; it had no compounds, no radicals, no look-alikes, no vocabulary and no mnemonics - and the compounds are the reason anyone looks a kanji up. It has all of them now, and it is built as an ordered list of blocks, each fed by exactly one source and each drawing nothing when it has nothing. The layering is the decision that shaped it: the dictionary sources are the root and WaniKani is the bonus. KANJIDIC answers for all ten thousand characters, the JLPT table adds the words for the two thousand the exams cover, and WaniKani adds its radicals, look-alikes, vocabulary and mnemonics for the two thousand it teaches. A character WaniKani has never heard of gets a complete page; one it teaches gets a better one. The word inside a compound is not a link - most are not WaniKani vocabulary and would land on a page saying nothing is there - but every kanji inside it is, and every chip goes through the same address function the search results use, so following a result and following a chip land in the same place. The word and radical pages took the same blocks on the way, so a related chip or a mnemonic reads the same on all three. The page stays the same for every reader: a member's own study state is a block for a later release, not a reason to make a shareable address vary by who opens it.

Area

study

Name

「うかぶうちがわ」 · 浮かぶ内側 Ukabu Uchigawa · the inside of a character, surfacing

Every list of characters here shows a different set of fields, and always did: the study queue has an SRS stage, the grade explorer has on and kun readings and no SRS at all, the JLPT explorer has a school grade and an N level. What went wrong was letting that difference decide the layout too, so five surfaces grew five sets of row markup and five ideas of where a reading goes. A list declares its columns now - a heading, a lane width, and how to draw one cell - and the shared component owns everything else: the single surface with hairlines between rows, the aligned lanes, a heading row that cannot drift from the cells beneath it because both read the same widths, the checkbox, the row button, the leading and trailing slots, and what folds away on a phone. The grade explorer shows item, meaning, on, kun, strokes and JLPT; the JLPT explorer shows item, reading, meaning, grade, JLPT, status and SRS; the study queue and history keep the WaniKani six. Different fields, one shape. The explorers' detail panel opens under the row that was clicked through a slot of its own, which was the last thing about their lists that no column could express and the reason each had kept private row markup for so long.

Area

study

Name

「いちらんのわく」 · 一覧の枠 Ichiran no Waku · one frame for every list

Three things a member could not do, all of them about lists. A list you saved yourself offered rename, edit characters, delete and practise, so the one thing you could not do with it was read it - the card's wall of glyphs was the whole view, with no meanings, no readings and no way into the glyph viewer. It opens now, into the same panel the Trouble and Favourites lists have always used, and the characters on the card open it as well as the Open action beside them. Bulk selection was reachable in the grid, where every card carries a checkbox, and unreachable in the list, where a row had none and clicking one opened the review modal instead of picking it - so a set could be gathered in one density and not the other, and Save to list sat above a list it could not be fed from. Rows carry the same checkbox now, and the same shift-click range. And the explorers drew their lists as forty bordered boxes stacked with gaps, each one an edge for the eye to stop at, while Study and History drew one surface with hairlines between the rows. Grades, the WaniKani library and the JLPT explorer take that chrome from the same place now.

Area

study

Name

「れつのなかみ」 · 列の中身 Retsu no Nakami · what is inside the row

The best list of subjects on the site was the one nothing could reuse. Choosing items in bulk opened a six-column table - item, reading, meaning, type, level, SRS - and it read better than every other list here, because a column of subjects is scanned one field at a time and that table was the only place the fields lined up. Everywhere else the reading sat squeezed under the meaning in a row that flowed to fit, so no two rows put anything in the same place. The table is gone and its shape is the shared list: study history, the study queue, the Trouble and Favourites lists and the bulk panel itself now draw the same lanes under the same headings, from one set of widths that the headings and the rows both read, so neither can drift from the other. On a phone the narrow lanes fold away and the reading returns under the meaning, because six columns do not fit in 393 pixels and a truncated reading is worse than a stacked one. Two things came with it: a row in the bulk list now removes its own item, where before picking one thing by mistake meant clearing all of them and starting over, and Select Visible works before anything is selected - it had been disabled until you had already chosen something by hand, which is the moment you least need it.

Area

study

Name

「よこならびのよみ」 · 横並びの読み Yokonarabi no Yomi · readings set side by side

Searching the WaniKani explorer printed the answer twice. A line above the results said Showing 5 search results and the line on the toolbar said Showing 5 of 5 items - the same number, two sentences, neither of them wrong, and no way to tell whether they were counting the same thing or two different things that happened to agree. There is one count now, and a search changes what it says it is counting rather than adding a second opinion beside it: forty of sixty-nine items when you are browsing, five of five search results when you are searching.

Area

study

Name

「ゆいいつのゆびさし」 · 唯一の指差し Yuiitsu no Yubisashi · one finger pointing, not two

Searching from the grade explorer threw away both the grade you were on and the thing you searched for. Standing on grade three and looking for a character put you on grade one with an empty search box. The grade had recently moved out of the query string and into the path, which is the right shape - the path says which collection you are looking at, the query says how you are reading it - and the reader that used to take a grade out of the query was removed along with it. What was not looked at was the search form, which went on submitting to the collection root with the grade as a hidden field, the shape from before. The root reads no grade any more; it only opens the first one. The form posts to the grade own address now and carries nothing but the search. A guard test holds the shape, because this is the failure mode of changing an address: the thing itself gets changed and one caller does not.

Area

study

Name

「ひきかえすひょうしき」 · 引き返す標識 Hikikaesu Hyoushiki · the sign that sends you back where you were

A saved list was fixed the moment it was made. Renaming arrived a while back, so the label could be corrected, but what was actually in the list could not - taking one character out meant going back to an explorer, finding the other forty, choosing them all again and saving over the name. In practice nobody edited a list; they built a second one and left the first sitting there. Every list now has an Edit characters control beside Rename and Delete. The characters become the control rather than growing one beside them, because they are already the biggest thing on the card: tap one to take it out. Adding is a field you type into, since somebody who wants a particular kanji can type it faster than they can find it in a picker, and what is typed goes through the same tidying a saved selection does, so a pasted sentence and a chosen set reduce to the same thing. The whole edit is one save at the end rather than a request per character, which is one chance to go wrong instead of six. Emptying a list is refused: that is deleting it, and deleting has its own button and its own confirmation. While the panel is open the card's own preview and count stand aside, so there are never two different numbers on one card with nothing to say which is being changed.

Area

study

Name

「ぬきさしのぬいめ」 · 抜き差しの縫い目 Nukisashi no Nuime · the seam you can unpick and restitch

The WaniKani explorer offered a tab reading All with the total beside it - two thousand nine hundred and twenty-two - and pressing it did nothing you could see. It put the member back on their own level, which was the level already showing, so the list did not change, the address kept the same single level in it, and the number on the tab described a view nothing in the app could open. There was no way to look across every level at all. It selects every level now and fetches the ones it does not already hold, a few at a time rather than all at once, because a level that is not cached goes out to WaniKani and sixty of those together is how an account gets rate limited. Results widen as each batch lands instead of waiting for the last. The address carries the whole selection, so the long view is a link somebody can be sent. The subtitle went with it: it read Select one level at a time, which was true of nothing - Sticky has always let a member gather several - and it was the last thing still saying one.

Area

study

Name

「ながいながめ」 · 長い眺め Nagai Nagame · the long view, every level at once

Choosing items on the study and WaniKani explorers could only ever be undone. Both run a bulk mode with shift-click ranges, and the panel counted what you had gathered and offered Select Visible, Clear and Done - nothing to actually do with the set. Meanwhile the grade and JLPT explorers, which choose through a different mechanism, had two destinations from the start: save it as a list, print it as a practice sheet. Those destinations belong to the act of choosing rather than to either mechanism, so they now live on the shared bulk panel and both explorers gain them at once. Practice takes only the kanji: a study queue is radicals and vocabulary too, and a practice sheet is squares to write kanji in - the list keeps everything chosen, and the sheet is withheld entirely when nothing chosen can go on one. Two smaller things had to move first. The save control took a selection object from one of the two mechanisms, which is what had kept it off the other; it takes characters now. And three surfaces worked out which member's page they were on by slicing the pathname by hand, which is one copy away from drifting - that is a shared hook.

Area

study

Name

「すすむすじ」 · 進む筋 Susumu Suji · a path that leads somewhere

A saved list could be created and deleted and nothing in between, so a typo in "Week 1" could only be fixed by throwing the list away and picking every character again. Lists now rename in place: the name becomes a field on the card, Enter saves it, Escape abandons it, and the characters are never touched. Renaming is its own request rather than a save under a new name, which matters more than it sounds - saving under a name that already exists replaces that list's contents, so a rename expressed that way would quietly empty whichever list already held the name. Choosing a name another list has now says so and keeps what was typed, because colliding with "Week 1" is an ordinary thing to do rather than an error to swallow. Names are tidied the same way whether saved or renamed, so "Week 1" and "Week 1" stay one list instead of becoming two nobody can tell apart. The two built-in lists, Trouble and Favourites, offer neither rename nor delete: they fill themselves as you tag while you study, and untagging the last item empties one rather than removing it. Three controls no longer fit a card at the width the grid gives it, so the row wraps rather than breaking "Practise these" across two lines inside its own button, and the condensed view stopped running off the right edge of a phone.

Area

study

Name

「かきかえるなまえ」 · 書き換える名前 Kakikaeru Namae · the name you rewrite

Trouble and Favourites arrived on the lists page in the last release and sat in the same grid as the saved lists, which quietly misrepresented both. They are permanent - they fill themselves as a member tags while studying, and there is nothing to delete because untagging the last item empties a list rather than removing it - while a saved list is made and thrown away at will. Run together, the grid invited a member to look for a delete on a card that does not have one. Each kind now has its own section under a heading that says what it is: Always here, which cannot be deleted, and Saved by you, which can. The card markup was extracted rather than duplicated per section, so both kinds still render through one component. The saved section says only what the page can actually do today - renaming a list is not built yet, and the copy does not pretend otherwise.

Area

study

Name

「るりのるいけい」 · 瑠璃の類型 Ruri no Ruikei · lapis, sorted by kind

Trouble and Favourites are lists in every sense a member cares about - they are named, they hold subjects, they are what you practise from - but they are tag rows rather than saved lists, so the page whose whole job is showing a member their lists was the one place they never appeared. They could be reached only through a panel opened by a button somewhere else, which meant the two lists a member had actually built were the two they could not find. Both now sit at the top of the lists page, in the same card as every saved list, with a count and a preview of what is in them. What differs is what they cannot do: there is nothing to delete, because untagging the last item empties a list rather than removing it, so the card offers Open instead and the panel is where tagging still happens. Their practice sheet is addressed by source rather than by the characters on the card, so printing one takes the whole list instead of the two dozen the card had room to preview. Both are always shown, empty ones included, because no favourites yet is an answer to the question somebody came to this page with and a missing card is not. The subtitle stopped claiming every list here was built by hand, which two of them now are not.

Area

study

Name

「らくだのらしんばん」 · 駱駝の羅針盤 Rakuda no Rashinban · the camel's compass

The study explorer had shift-click already, and had it wrong in the same way the WaniKani explorer did: the anchor a range was measured from was the item is position in the filtered list, which stops meaning anything the moment a filter changes or another page loads. Change a status chip after picking one item and the next shift-click swept a stretch the member never crossed. It anchors on the subject now, which survives both, and it uses the same range helper the other four surfaces use rather than a third copy of the arithmetic. That is all of them: the three explorers, study history and the tagged lists now sweep a range the same way, and a test names each surface so a sixth cannot quietly go without.

Area

study

Name

「よそいのよはく」 · 装いの余白 Yosoi no Yohaku · room left in the arrangement

Choosing reached the explorers in 0.168.0 and stopped there, which left the two surfaces where a member has already gathered something deliberately - their study history, and the Trouble and Favourites lists - with no way to act on a set. Scanning what you got wrong last week is exactly the position from which you want those characters on a practice sheet, and the only route was to write them down. Both now offer the same Choose mode, in both densities, with shift taking everything between the last pick and this one. The work went into the shared row and card pair rather than into either surface, so neither had to learn anything: they pass the selection through and get the tick, the highlight and the sweep for free, and the next list built on that pair inherits all of it. The tick sits before history is own correct or wrong mark rather than replacing it, because whether you got it right is exactly what you are reading when you decide whether to pick that one. Where a surface cannot know which member it belongs to - the tagged lists open from the game lobby, from history, from either explorer - the practice offer is withheld rather than pointed at a guess.

Area

study

Name

「ゆたかなゆびわ」 · 豊かな指輪 Yutaka na Yubiwa · a ring with plenty on it

Choosing a set of characters existed in the grade explorer and nowhere else, and the JLPT explorer had no way to select anything at all - so building a practice sheet or a saved list from a JLPT level meant picking the kanji out by hand somewhere else. The JLPT explorer now offers the same Choose mode, the same bar, and the same two destinations, and shift-click takes everything between the last card and this one. None of that is a second copy: the bar and its destinations moved into a shared component that the grade board now uses too, and the card draws the ring and the tick itself so a chosen card reads the same on every surface. The tick sits where the index number was, because every corner of the card is already spoken for by a pill, a level, a success rate or the trouble and favourite controls, and a tick laid over any of them hides something the member is using to choose. The WaniKani explorer keeps its own selection - it picks subject ids and carries bulk operations the others have no use for - but it no longer carries its own copy of the range maths, and that fixed a bug in passing: its anchor was the item position in the visible list, so changing a filter or loading another page moved every position underneath it and the next shift-click swept the wrong stretch. It anchors on the subject now, which survives both.

Area

study

Name

「やまもりのやさい」 · 山盛りの野菜 Yamamori no Yasai · a heaped plate of vegetables

In the grade explorer s list view the stroke-order button was positioned absolutely at the right-hand edge of each row and revealed on hover, so it came down on top of the JLPT pill it was meant to sit beside. The row had reserved padding for it, but the pills spread into that space anyway and the two ended up in the same place - reserving room was never going to be the fix, because a row s controls have to take part in the layout. It sits in the line now, as the last thing in the row s chip cluster, taking its own room rather than floating over whatever is underneath. The card view is unchanged: there the button has a free corner to occupy, and one permanent control on every card of a screenful is noise rather than help, so it stays where it was and still waits for hover.

Area

study

Name

「むりのないむぎ」 · 無理のない麦 Muri no nai Mugi · barley that is not forced

The JLPT and WaniKani explorers share one card component, and its list view was unusable: where each row s character should have been there was a pile of chips instead - the level pill sitting on the success rate sitting on the glyph. Both explorers hand the card an overlay written for the card shape, which places the level, the success rate and the trouble and favourite controls in the corners of a glyph box 128 pixels tall. A row s glyph box is 44 pixels square and has no corners to spare, so everything landed on top of everything else. The card now tells its overlays which shape they are being drawn into: in a row they lay themselves out in the line as ordinary chips and buttons, while the card keeps the corners it was designed around. Neither explorer changed - they pass the same overlay as before - and the trouble and favourite controls now work from the list view as well as the grid, which is what a member scanning a long level actually wants. Measured on both explorers afterwards: forty row glyph boxes each, none holding anything but the character.

Area

study

Name

「まっすぐなまち」 · 真っ直ぐな街 Massugu na Machi · a street set straight

Choosing characters was one click per character, so picking a row of five in the grade explorer took five clicks and picking forty took forty. Shift-click now takes everything between the last card and this one, the way it works in a file browser or a mail client, in either direction and from either end. The range only ever adds - a sweep that unpicked half of what it crossed would be impossible to predict from where it began - and the far end becomes the new anchor, so a selection can be walked outward a stretch at a time. A keyboard user gets it without a second set of handlers: activating a card with Enter or Space reports the modifiers held at the time, so Shift and Enter sweeps the same range. The logic sits in the shared selection hook rather than in the grade grid, so the next surface to offer choosing inherits it, and the hint appears in the selection bar only once something is picked, which is the first moment a range is possible.

Area

study

Name

「はこぶはこ」 · 運ぶ箱 Hakobu Hako · the box that carries

In the grade explorer's list view, a kanji with many kun readings ran its readings straight out through the right-hand edge of its row. 下 has eight of them and they crossed the card border, the JLPT pill and whatever sat beside it. The text had been asked to truncate, but a flex item's floor is the width of its own content, so the paragraph holding the readings simply sat as wide as they measured and the truncation inside it never had a narrower box to work against. Measured after the fix at 393, 650 and 1440 pixels: no row overflows its card by a single pixel, and the page has no horizontal scroll.

Area

study

Name

「ねばるねずみ」 · 粘る鼠 Nebaru Nezumi · the mouse that holds its ground

The writing practice page was built as a print preview and painted like one: white background, neutral greys, black text, from the source chips at the top to the squares at the bottom. It was the one page in the app that looked like a photocopy of the rest of it. Everything on screen now uses the site's own palette - the chips take the accent, the sheet sits on a surface card the way every other list does, the guide squares are drawn in the line colour and the tracing strokes in a pale accent, which is what a Japanese workbook prints them in anyway. None of it reaches paper: every colour on a printed element states its printed grey beside itself, the sheet's card drops its border and padding so the squares start at the printer's margin, and a rule under the whole region strips any background a later panel might bring with it. A test reads the sheet's own source and fails on a colour that says nothing about how it prints.

Area

study

Name

「ぬりかえたぬの」 · 塗り替えた布 Nurikaeta Nuno · the cloth, dyed again

The previous release protected what a page draws on arrival and nothing else, so the stroke-order modal still showed "circle" where 円 belonged in both its Gothic and Mincho previews, and the shared explorer card lost its protection the moment a list switched to rows. Those and the game's match prompt are covered now, along with the admin kanji tables. A second half to it: translation was also respacing compact chips, because a label built from several JSX children is several text nodes with the gaps between them up for grabs - "JLPT N5" came back as "JLPTN5", "APPR - SRS 4" as "APPR- SRS4", and every count as "( 157 )". Codes and counts are now single strings that refuse translation, while prose like "Showing 1-60 of 80 kanji" still reads in whatever language the reader wants. A source guard fails the build on any element that reaches for a Japanese typeface without refusing translation, which is what would have caught the modal.

Area

study

Name

「けんごなけっかい」 · 堅固な結界 Kengo na Kekkai · a solid warding barrier

Chrome's translate offer used to eat the site. On the grades explorer it replaced every kanji with its English meaning - 一 became "one", 右 became "right" - beside a card already printing that same meaning underneath, and it romanised the readings, so a page teaching kanji showed no kanji at all. Every Japanese element now carries both refusals, the HTML5 translate attribute and Google's notranslate class, while the English chrome around it stays translatable for anyone who needs it. The font class and the glyph sizing helper each carry the marker with them, since anything drawn in the Japanese face is Japanese, and a unit test keeps the raw font class out of the rest of the tree so the next glyph surface cannot take the styling and forget the protection. Release notes that quote Japanese inside an English sentence mark only the quote. Audited across sixteen pages: no unprotected Japanese left.

Area

study

Name

「くずれぬくら」 · 崩れぬ蔵 Kuzurenu Kura · the storehouse that does not fall

A screen page is three sheets of paper now, so printing one comes out whole. It was twenty characters at every square size - a number about the screen, not about paper - and twenty medium squares is two and a half sheets, so the third sheet of every page came out 40% blank. The counts were measured by adding characters one at a time and watching where the paper broke: seventeen large, twenty-two medium, twenty-six small, each landing on a US Letter sheet boundary and taking the same three sheets on A4. Rows are numbered as well, by their place in the whole list rather than the page, so page two runs 23 to 44 and a sheet can be set as numbers twelve to twenty.

Area

study

Name

「かみのかず」 · 紙の数 Kami no Kazu · the number of sheets

Print asks how much: this page, or everything. It printed the twenty characters on screen and nothing else, so a grade of eighty was four separate print jobs - and because twenty characters is about two and a half sheets of paper, three of those four ended in a mostly-empty page. Everything drops the reading pages and lets the characters flow, which takes grade three from ten jobs and thirty sheets to one job and twenty-six. Lists too long for one document split into print runs rather than rendering 1,110 characters at once. Where the pager sits is no longer a question the sheet asks.

Area

study

Name

「おおばんおてほん」 · 大判お手本 Ooban Otehon · large-format copybook

Choose characters on the grades explorer, give the set a name, and it keeps: Week 1, Week 2. A Lists page shows every set you have built with the characters it holds rather than a count, and each one goes straight to a practice sheet. The same idea as Trouble and Favourites, with a name instead of a flag.

Area

study

Name

「あしあとあやとり」 · 足跡あやとり Ashiato Ayatori · footprints in a cat's cradle

Choosing items is a surface control now, sitting beside the grid and list toggle rather than being a feature of the practice page. Pick characters on the grades explorer and send them to a sheet; the chosen set travels in the link, so a sheet built by hand is shareable and printable before anything is saved. Destinations are offered to a chosen set rather than built into the picking.

Area

study

Name

「わかばわだち」 · 若葉轍 Wakaba Wadachi · new leaves in the wheel ruts

The sheet named Letter as its paper size, which overrode the printer: an A4 tray produced a Letter page scaled down inside it. It now pins only the orientation and lets the paper be whatever is loaded, so A4 and 8.5x11 both come out filling the page.

Area

study

Name

「ろせんろくおん」 · 路線録音 Rosen Rokuon · a recording of the route

The writing sheet offers three square sizes. Large gives a child six big boxes to form characters in; Small gives an adult ten smaller ones and an extra tracing repetition, which is what copying at volume actually wants. Medium is the practice-book square the sheet always had.

Area

study

Name

「れんぞくれきし」 · 連続歴史 Renzoku Rekishi · history, without a gap

Printing a writing sheet now gives you a worksheet: Letter with proper margins, a Name and Date rule at the top, characters that never split across a page break, and none of the site's footer or version banner. The faint tracing characters print faintly instead of vanishing into the browser's ink saving.

Area

study

Name

「るろうるいせき」 · 流浪累積 Rurou Ruiseki · what wandering piles up

The explorer and the writing sheet both offer the four levels the test ran until 2009, for anyone who sat it then. Both schemes count down, so old Level 4 is the beginner paper and matches N5 rather than N4 - and N3 is absent, because it was added in 2010 to bridge old Levels 3 and 2. The explorer says so rather than inventing a mapping.

Area

study

Name

「まどのまぼろし」 · 窓の幻 Mado no Maboroshi · a mirage at the window

The option reached the stroke-order sheet but not the tracing sheet, which kept printing the character solid in the first square - the one thing the option exists to remove. It is now an empty square to write in, like the rest of the row.

Area

study

Name

「ほおずきほそみち」 · 鬼灯細道 Hoozuki Hosomichi · a lane of lantern plants

Filters and results shared one card on the grade explorer, divided by a hairline, while both other explorers keep them in separate cards. They are separate here too now, and the result count sits with the results rather than with the filters.

Area

study

Name

「へやのへきが」 · 部屋の壁画 Heya no Hekiga · a mural in the room

Trouble and Favourites are now sources for a writing sheet, offered from the practice page itself rather than only where the tags were made. They stay quiet until the control row is hovered, and appear always on a touch screen where there is no hover to give.

Area

study

Name

「ふぶきのふたば」 · 吹雪の二葉 Fubuki no Futaba · two leaves in a snowstorm

Whether the finished character sits in the first column is now a checkbox rather than a decision made for you, and the readings can print beside each meaning. The chooser also closes when you pick what is already picked, or with a Done button at the end of the row.

Area

study

Name

「ひとやすみひだまり」 · 一休み日だまり Hitoyasumi Hidamari · a rest in a patch of sun

The WaniKani and JLPT explorers can be read as rows as well as cards, which the grade explorer already offered and they did not. The four pages behind the Explore tabs also shared no wrapper, so the sub-nav sat at three different heights and moving between tabs nudged the whole header; they now agree exactly.

Area

study

Name

「はつゆきひろば」 · 初雪広場 Hatsuyuki Hiroba · first snow on the square

Selecting the source you are already on opens a second row holding every value it offers: the eight school grades, sixty WaniKani levels or five JLPT levels. Changing sheet meant going back to the grade page and in again, which made a page you might print several of from harder to use than it needed to be.

Area

study

Name

「のはらののぞみ」 · 野原の望み Nohara no Nozomi · a wish across the meadow

Every row is a full eight squares now: the finished character faintly at the start to work from, the strokes after it, and empty squares filling out the rest of the line. Each row is also titled with the character itself before its meaning.

Area

study

Name

「ねがいのねいろ」 · 願いの音色 Negai no Neiro · the tone of a wish

Changing grade or turning the page dropped you back to the tracing sheet, because only the sheet toggle carried the choice. Every link on the page keeps it now.

Area

study

Name

「ぬけみちぬま」 · 抜け道沼 Nukemichi Numa · a shortcut past the marsh

Writing practice can now be printed as a stroke-order chart: one numbered square per stroke, the character built up to that point with the newest stroke picked out. A character of more than eight strokes runs onto a second row instead of shrinking the squares. On a phone the sheet now says plainly that it wants a bigger screen, rather than rendering squares too small to write inside.

Area

study

Name

「にじのにわ」 · 虹の庭 Niji no Niwa · a garden under a rainbow

The stroke-order panel is now a shared component any surface can open, and every kanji has a public page you can send to someone. Selecting a card in the grade explorer opens it, which is how every other subject grid already behaved. The header carries the readings as well as the meaning, since a reader knows 何 by なに before they know it by what, with the stroke count beside it and the credit centred at the foot.

Area

study

Name

「とうげのともしび」 · 峠の灯 Toge no Tomoshibi · a light on the mountain pass

The grade explorer can be read as rows as well as cards, which is how you scan a whole school year and the only way to see a kanji with ten kun readings in full. The stroke control no longer sits on every card: it appears when you hover, or always where there is no hover to give, which also fixes it covering the end of a long reading.

Area

study

Name

「ちどりちゃや」 · 千鳥茶屋 Chidori Chaya · a teahouse where the plovers gather

The Reviews menu was anchored to its right edge to keep it on screen on a desktop, which pushed it off the left edge on a phone, where the panel is nearly as wide as the display. It now anchors left on small screens and flips right only where there is room.

Area

study

Name

「けいこびより」 · 稽古日和 Keiko Biyori · a fine day to practise

Writing practice is no longer only for school grades: the same sheet prints a WaniKani level or a JLPT level, so the ladder you are actually climbing is the one you can practise writing.

Area

study

Name

「おりがみおはぎ」 · 折り紙おはぎ Origami Ohagi · origami rice cake

Stroke order moves into a shared button any surface can drop in, and the glyph viewer gains it for kanji. The data now covers every WaniKani kanji as well as the school grades - the characters WaniKani teaches past the joyo list were silently missing, and those are the ones a learner is least sure how to write.

Area

study

Name

「ろじうらろくろ」 · 路地裏轆轤 Rojiura Rokuro · back-alley potter's wheel

Reported by John: the level read as a pill in the lists and as faint borderless text on the study card and in the view modal. The in-glyph level and success rate are pills everywhere now.

Area

study

Name

「めぐみめぐり」 · 恵み巡り Megumi Meguri · blessing pilgrimage

Subject cards put the level in the top-right corner the way every other surface does, and the remove button moves to the bottom-right instead of squatting where the level belongs.

Area

study

Name

「はるかぜはなみ」 · 春風花見 Harukaze Hanami · spring-breeze blossom viewing

The Study queue offers the same grid/list toggle the tagged lists and history already use, so a long queue can be scanned one line at a time. The choice is remembered per surface.

Area

study

Name

「のどかのれん」 · 長閑暖簾 Nodoka Noren · tranquil shop curtain

The queue tag filter now shows as a chip you can clear, so a sticky Trouble filter no longer hides most of the due reviews without saying so, and the Trouble & favourites lists open from the Study page rather than only the game lobby, History and the JLPT explorer.

Area

study

Name

「なつまつりなると」 · 夏祭り鳴門 Natsumatsuri Naruto · summer-festival whirlpools

Read subject content locally: the study page went from 15.2s to under a second warm.

Area

study

Name

「えにしのえま」 · 縁の絵馬 Enishi no Ema · votive plaque of destiny

Tap a status to see its stages; filtered totals stay honest.

Area

study

Name

「よあけようかい」 · 夜明け妖怪 Yoake Youkai · daybreak yokai

Compare two items during a review session.

Area

study

Name

「もちもちもみじ」 · もちもち紅葉 Mochimochi Momiji · springy autumn maple

Advance the queue immediately and persist history behind it.

Area

study

Name

「めでたいめろんぱん」 · めでたいメロンパン Medetai Melonpan · auspicious melon bread

One status chip row shared by review and detail modals.

Area

study

Name

「みどりみそ」 · 緑味噌 Midori Miso · green miso

Reveal a meaning without leaving the review.

Area

study

Name

「ほかほかほたる」 · ほかほか蛍 Hokahoka Hotaru · toasty firefly

Rank reviews by measured ease so hard items surface first.

Area

study

Name

「ふわふわふじ」 · ふわふわ富士 Fuwafuwa Fuji · fluffy Mount Fuji

A faster grading loop for large queues.

Area

study

Name

「ひかるひつじ」 · 光る羊 Hikaru Hitsuji · glowing sheep

Every filter chip carries its own count; sort by time waiting.

Area

study

Name

「ぬくもりぬりかべ」 · ぬくもり塗壁 Nukumori Nurikabe · warm-hearted wall yokai

Show the stage an answer moved an item to.

Area

study

Name

「とびきりとんかつ」 Tobikiri Tonkatsu · exceptional pork cutlet

SRS bucket filters and an in-modal item view.

Area

study

Name

「ちいさなちょうちん」 · 小さな提灯 Chiisana Chouchin · little paper lantern

Item-spread drilldowns across the last five levels.

Area

study

Name

「たのしいたぬき」 · 楽しい狸 Tanoshii Tanuki · merry tanuki

Multi-select reset with staged confirmations.

Area

study

Name

「せっかちなせんべい」 · せっかちな煎餅 Sekkachi na Senbei · impatient rice cracker

Hide metadata until reveal, then grade yourself.

Area

study

Name

「さくらざむらい」 · 桜侍 Sakura Samurai · cherry-blossom samurai

Self-graded attempts recorded locally, independent of WaniKani.

Area

study

Name

「けだかいけんだま」 · 気高いけん玉 Kedakai Kendama · noble kendama

Level, type and locked filters that survive a reload.

Area

study

Name

「くろいくま」 · 黒い熊 Kuroi Kuma · black bear

Two distinct flows rather than one queue doing both jobs.

Area

study

Name

「きらきらきつね」 · きらきら狐 Kirakira Kitsune · sparkling fox

One item at a time with WASD and arrow navigation.

Area

study

Name

「かがやくかっぱ」 · 輝く河童 Kagayaku Kappa · gleaming kappa

Browse subjects with linked kanji relations, readings and radical meanings.

Area

study

Name

「いきないざかや」 · 粋な居酒屋 Iki na Izakaya · stylish pub