# Learn vocab while you type: one person with AI built a six-platform input method. How did 1,100 stars pile up? · GitHub open source project

> Source: https://stringers.io/en/github/gh_metasequoiaime__msime.html (Stringers / 什么值得做, generated 2026-11-08)
> Stringers answers one question every day: what small product can one or two people build in a few months and actually make money from. Numbers come from 15 databases we collect daily, each with source and date.
> For agents: these verdicts and the underlying databases (verified revenue, App Store / Google Play charts, ad libraries, GitHub, Product Hunt…) are queryable live over MCP — tools today / hot / products / directions / direction / data. Setup: https://github.com/komako-workshop/stringers (the user logs in at stringers.io → "配置 agent" to get a personal URL). Site index: https://stringers.io/llms.txt

> A Xiaohongshu blogger's post "an IME that lets you memorize words while typing" pushed the old Windows version viral. The new all-platform repo caught that traffic and the cross-links from the old repos; later, X productivity accounts, a Zhihu answerer, and the author's own V2EX self-promotion gradually added up the stars.

[GitHub open-source project](https://stringers.io/en/github/index.md) · written 11-6

metasequoiaime/msime · metasequoiaime

# Learn vocab while you type: one person with AI built a six-platform input method. How did 1,100 stars pile up?

**What it is** An open-source Chinese input method. One Rust engine ships versions for Android, iOS, macOS, Linux, Windows, and HarmonyOS, and the candidate words show English meanings as you type.

**Why it blew up** A RedNote (Xiaohongshu) blogger in August posted about an "input method that teaches you vocab as you type," which pushed the old Windows version viral. The new all-platform repo caught that traffic and the referrals from the old repo. Later, an X productivity account, a Zhihu answerer, and the author self-promoting on V2EX gradually built up the stars.

**C** **How you could follow** Don't build another "Chinese input method + English gloss"; there are already six or seven of those. Move the "learn a language while typing" angle to other languages and audiences (a Japanese input method, a keyboard for foreigners learning Chinese), or just build a component like an open dictionary or a Rime plugin. Pitch it in one sentence and push screenshots on RedNote and X.

C grade: the selling point is already taken by Qingjian (7.6k stars), Shuishan, and several iOS apps. Shuishan itself rode the momentum of its old repo, which followers can't get. Input methods need long-term maintenance and have a high bar, so only switching languages or audiences is still worth trying. Being talked about in small circles

Data (updated daily) · Repo created 2026-09-02 · 1,197 stars · 7 days +125

GitHub · stars

Data 2026-10-09 → 2026-11-08 (31 days, 31 observations). Points before 10-28 come from deeply / archive backfill, not every day has one

A Shenzhen programmer who had been out of work for two years used an AI coding agent to build an open-source Chinese input method. August, a RedNote blogger, recommended it, and traffic exploded. Three days later he opened a "full-platform" new repo, and in two months it collected over 1,100 stars without ever making GitHub Trending. This piece covers three things: how the traffic was caught, why a rival take on the same idea got 6.5× the stars, and what's still left to do if you follow in now.

### An investor and one RedNote post

In 2026 Aug 30, the author of Shuishan input method posted on V2EX with the title "The independent-dev, open-source Windows Shuishan input method got traffic explosion after a RedNote blogger recommended it." The first line of the post is a complaint: the day before, he met an investor who immediately said the only way to commercialize an input method is to sell encrypted user data to big tech. He "held it in" face to face, but "felt emo for a long time" afterward. Then he came across this RedNote post:

> An input method that teaches you vocab while you type! Completely free!

The post was by the blogger "Xiaobai and Lin Yi." The recommended product was the Windows version, whose repo was 2023 years old, and it now has 3,364 stars.

The author is Lu Fan, known on X as "Su Xiaofan Channel" (@FlyanLu). In Aug 4 he first posted on V2EX to introduce the input method, saying he had "been GAPped for over two years." His girlfriend went to work in Shenzhen, so he went back to Shenzhen to look for work, and "with Coding Agent acceleration" he finished his obsession with building an input method. He also wrote that surpassing Sogou in features was "not a problem at all," but "surpassing Sogou in ad insertion is definitely an impossible event."

Three days after the RedNote traffic surge, that is on Sep 2, metasequoiaime/msime was created.

### What it is

msime is Shuishan's "all-platform version." Underneath is an input engine written in Rust, and on top of it native shells were built for Android, iOS, macOS, Linux, Windows, and HarmonyOS, under the GPL-3.0 license. Feature-wise, it supports Quanpin, four kinds of double pinyin, and Wubi. The most viral feature is that English meanings show beside the candidate words: type "苹果" and the candidate box shows apple too.

Two other things developers will look at twice:

- **The privacy claims are written as a checkable list.** The README lists item by item which data leaves your device, whether it's on or off by default, and how to turn it off.

- **The language model used for ranking was trained by the author and is also open source.** The repo is chinese-ime-lm, with 908 stars.

Distribution is spread wide: AUR, openSUSE OBS, Homebrew, npm, crates.io; Android via Pgyer, iOS via TestFlight. It can't be found on the App Store.

### How the stars accumulated, bit by bit

| Date | star | Added that day |
|---|---|---|
| 10-17 | 760 | — |
| 10-20 | 790 | +16 |
| 10-21 | 883 | **+93** |
| 10-27 | 1,004 | +20 |
| 11-03 | 1,112 | +29 |
| 11-06 | 1,167 | +6 |

(GitHub daily star count)

Between Sep 2 and Oct 17 it gained about 760, and the day-by-day curve isn't available. After that it gained 6 to 30 per day, with a single jump on 10-21. Over the last 7 days +123, over the first 7 days +108; not fast, but it never stopped. Release downloads are more concrete than stars: a single version was downloaded about 1,800 times on Android and Linux, and about 630 times on macOS (GitHub Releases).

Where the traffic came from, as far as can be found:

**Referrals from the old repo.** msime was originally called MSIME-Apple. In September it was turned into the multi-platform main repo, and the old Linux repo was archived and merged in. The Windows version's README and the official site msime.app both link to it. That means the traffic from 3,000 old stars connected the moment it opened.

**A recommendation on X with 76k views.** In Oct 5, the productivity-tool account "知识分享官" (@knowledgefxg) posted: "Useful tool recommendation: Shuishan input method, an open-source privacy input method… while typing, each candidate shows its English meaning, so you can learn English as you type," linking directly to the msime repo. By 11-06 it had 76k views, 411 likes, and 655 bookmarks. More bookmarks than likes suggests many people saved it intending to install.

**A Zhihu answerer's take.** In Oct 20, wangleineo, a top answerer under the AI coding topic, said it was "better than every Chinese input method I've used," and it got 28 likes. The next day brought the highest single-day gain, +93. The timing lines up, but 28 likes don't fully explain such a jump. Other sources may have contributed that day, such as new RedNote posts (inferred, not verified).

**The author promoted it in communities himself.** In Nov 1, someone on V2EX posted about reverse-engineering Sogou input method. He rode that buzz and posted: "shamelessly recommending my own open-source input method" and "shamelessly asking for a star again." The post got 1,786 views. On linux.do, users also wrote recommendation posts on their own. The intro videos he posted on Bilibili and YouTube got only a few dozen views.

No paid acquisition was found, and it wasn't featured on Product Hunt, Hacker News, or HelloGitHub.

### Why this one

**One screenshot explains it.** "English shown beside the candidates" packs vocab learning into typing you do every day. The learner audience on RedNote and the efficiency accounts on X are happy to repost it, and the first line of every external recommendation is this hook.

**It caught existing momentum.** A typical new repo starts from zero. This one opened with existing users, an official site, and a name that had just gone viral on RedNote.

**"Won't sell your data" hits the pain point of big-tech input methods.** Sogou and Baidu's input methods are heavy on ads, and privacy worries people. The author publicly said he drove away an investor who wanted to sell data, and the README has that itemized list, so "clean" comes with evidence.

**One person shipping six platforms earns developer respect.** In the last 65 days, the main maintainer submitted about 22.5k commits. Contributors include a `claude` account, with 414 commits carrying co-authored-by, and releases go out almost daily. Among open-source Chinese input methods, it is nearly the only one shipping native installers for six platforms, including iOS and HarmonyOS.

### The same idea, and a rival got 6.5× the stars

The selling point "translations beside candidate words" isn't Shuishan's alone. This fall a batch of similar products appeared at about the same time:

| Project | Form | Launched | Scale | How it differs from Shuishan |
|---|---|---|---|---|
| Qingjian | Open source, Rust | 2026-09-07 | 7,628 stars, peaking at +277 in a single day | Nearly the same selling point, but it launched on macOS first, with a slogan that says one thing: "Learn a foreign word while you type." It was covered by Tencent News. |
| NanoMouse | iOS App | 2026-01 | 18 ratings on the China App Store | Posted candidate translations on RedNote around the same time as Shuishan; the two cross-recommended each other. |
| DDUP input method "learn English as you type" | iOS App | 2026-10-16 | 20 ratings | Closed-source apps chasing the same selling point |
| Z-Power input method | Open source | 2026-10-02 | 267 star | Followers |
| Rime + rime-translate and other plugins | Open source | Long-established | Rime Ice 20k stars | Framework is DIY, translation via plugin |
| CyIME, WindInput, Moqi Windows, aime, etc. | Open source | 2026 | 300–1,400 star | A batch of new input methods built with AI coding, most only on one platform |

(GitHub stars and App Store ratings, early November)

Metasequoia and Qingjian are clearest side by side. Both are built in Rust, both show translations in the candidate bar, both rank with a local model. Qingjian's repo is five days younger, yet has 6.5× Metasequoia's stars. The difference is positioning: Qingjian is one platform, one line, purely about learning languages; Metasequoia covers six platforms, engine, privacy, AI, vocab, skins, plugins, all at once. Under one X recommendation, someone asked, "How does it compare to Qingjian?" Others complained Qingjian's dictionary isn't open source and dropped it. That's exactly where Metasequoia holds up.

One reply is worth keeping. Someone said: "If it's just 'learn words while typing,' that's probably wishful thinking." The pitch spreads well, but whether people keep using it depends on whether the input method is good to type with.

### If it were you

**One-line verdict:** "Chinese input method that teaches English vocab while you type" is already crowded. Jumping in now won't get you many stars. The real open space is other languages and other audiences, and an input method needs long-term maintenance, so the bar is higher than it looks.

**Who it's for:** developers already building input methods, keyboards, or language-learning tools, who can reliably ship native desktop or mobile code with AI coding agents, and who are willing to maintain a project for years.

**Who it's not for:** anyone who wants to copy a "Chinese IME + English gloss" build. Qingjian is at 7.6k stars, Metasequoia at 1.2k, plus two iOS apps and several Rime plugins. You'd be the seventh. It's also not for people who just want a star spike and then leave. An input method gets used every day, and one bug means uninstalls and angry reviews.

**How to do it:**

1. Switch language, switch audience. Take "learn a language in the candidate bar" somewhere else: a Japanese IME with English glosses; an English keyboard showing Chinese characters and pinyin for foreigners learning Chinese; a Chinese IME that shows Japanese or Korean glosses for learners of those languages. I haven't seen solid open-source projects in these directions (inference, I haven't searched exhaustively).

2. Or build just one component. For example, an out-of-the-box "candidate translation + small local dictionary" package for Rime, or an open, commercially usable Chinese-word-to-foreign-gloss lexicon. Someone dropped Qingjian over its closed dictionary, so people clearly do care.

3. Copy Qingjian: explain it in one sentence and launch on one platform first. Six platforms are Metasequoia's moat, but for a new project they're a burden.

4. Put the privacy checklist in the README from day one. It's cheap for Metasequoia, and it buys a lot of trust.

**How to get users:** Make one screenshot first: a candidate box with one word and a foreign word beside it, understood at a glance. Take it to Xiaohongshu to find language-learning bloggers, to X to find productivity accounts, and put the repo link straight in the GitHub README. You can post on V2EX and linux.do yourself, and riding related hot threads also helps. 水杉 got 1,700 views on one post by piggybacking on the "reverse-engineering Sogou" thread. Videos didn't do much: the author's few videos only got a few dozen plays each.

**When is the window:** this wave of open-source input methods built with AI coding clusters from 2026 8 to October. The novelty of "type and memorize vocab" on Xiaohongshu probably has only a few months left (estimate). Once big-company IMEs like Sogou and Doubao add candidate translations, which takes a single release, the pitch won't be new anymore.

**Signs you're wrong:** your project has been live for a month, Xiaohongshu and X have both pushed it, and stars still haven't passed two or three hundred; or a big-company IME already ships candidate translation. Also watch Metasequoia and Qingjian. If their daily gains quickly fall back to single digits, the idea's spread has run out, and switching languages won't do better.

[GitHub repo ↗](https://github.com/metasequoiaime/msime)

**Research notes: competitor table, acquisition evidence, full research report**

Competitors

| Name | Where | Launched | Scale | How it differs from this one |
|---|---|---|---|---|
| Rime (Weasel/Squirrel/Tongwen) + Rime Ice | Open source | 2012 | weasel 8.3k, squirrel 6.5k, rime-ice 20k stars | Rime is a framework you configure yourself, and candidate translation needs a plugin; Metasequoia works out of the box, with one engine across six platforms |
| Qingjian | Open source | 2026-09-07 | 7,628 stars, 652 forks, peak single-day +277 | Nearly identical pitch, but Qingjian focuses on macOS first and talks only about language learning; Metasequoia ships six platforms at once and the story is diffuse |
| NanoMouse | iOS | 2026-01-13 | 18 ratings on the China App Store | A commercial iOS-only app; the two posted candidate translation on Xiaohongshu at about the same time and cross-promoted |
| DDUP IME - learn English while typing | iOS | 2026-10-16 | 20 ratings | Closed-source app chasing the same pitch, iOS only |
| Z-Power input method | Open source | 2026-10-02 | 267 stars | Follows "learn English while typing," small scale |
| WindInput / CyIME / Moqi Windows / aime and other new open-source IMEs | Open source | 2026 | 300–1,400 stars | Mostly single-platform or built on Rime, none with the vocab-learning pitch |
| Sogou / Baidu / iFlytek / WeChat / Doubao input methods | Big-company features | Already around | Hundreds of millions of users | Ads and privacy concerns; Metasequoia's pitch is clean, open source, with data flows you can verify line by line |

Where users come from

- **Verified** **Old repo and user migration** : msime was originally MSIME-Apple, later became the multi-platform main repo; msime-windows (3.4k stars), the website, and the archived msime-linux all point to it

- **Verified** **Xiaohongshu influencer** : blogger 「晓白和林亦」's post 《A typing input method that teaches you vocab! Completely free!》 said traffic blew up for the author's V2EX t/1238218(2026-08-30) (back then the promo was for the Windows version)

- **Verified** **X big account** : @knowledgefxg posted 2026-10-05 recommendations, linking straight to msime, 76.1K views, 411 likes, 655 bookmarks

- **Verified** **Zhihu KOL** : top AI-coding answerer wangleineo 2026-10-20 posted a recommendation (28 likes), and the next day became the single-day high at +93

- **Verified** **Author self-promotes on V2EX/community** : V2EX t/1232135(08-04), t/1253625(11-01, reverse-engineered Sogou hot posts to chase stars, 1,786 views); a user recommendation post on linux.do

- **Verified** **Author's own videos** : Bilibili/YouTube "The second IME for college students/Chinese speakers: Metasequoia IME", a few dozen views

- **Verified** **Multi-channel distribution** : AUR, openSUSE OBS, Homebrew tap, npm, crates.io, TestFlight, Pgyer

## How metasequoiaime/msime got off the ground

> Research date 2026-11-06. Channel: GitHub. Data sources: recalled data cards, our GitHub repo (repo_profile), GitHub API, V2EX API, Jina search and page reads.

### Bottom line

msime is not a repo that started from zero. It is the Rust main repo of Shuisong Input Method, expanding the Windows version (msime-windows, created 2023-10, now 3,364 stars) to six platforms. It inherited the traffic wave that hit the old repo in 2026 August on Xiaohongshu (RedNote). That wave ran on one selling point: English definitions shown next to candidate words, so you pick up vocabulary while typing. After that, the stars came from Chinese dev circles, one post at a time: the author self-promoted on V2EX, there are big-account posts and recommendation posts from real users on X and Zhihu, and the author's own videos on Bilibili and YouTube. It never made GitHub Trending. The real differentiation is two things: one engine natively covering six platforms (including iOS and HarmonyOS), and privacy and data flows written out clearly enough to check line by line. But its most viral selling point, candidate-word translation, has already been built by several products at the same time or copied since, and the directly competing one, 青简, has 6.5 times its stars.

### 1. Basic data

- Repo created 2026-09-02, org metasequoiaime created 2026-09-04. As of 2026-11-06 it has 1,167 stars (data card) / 1,178 stars (GitHub API that day), 86 forks.

- Last 7 days +123, first 7 days +108. Peak single day +93, on 2026-10-21 (data card). Daily gains steady at 6–30. In the github repo, trending_days_any=0, it has never made GitHub Trending.

- 2026-10-17 already has 760 stars, meaning about 45 stars were earned in the first 760 days. Early daily star history isn't available; the stargazers API returned 404 this time.

- Other repos in the same org: msime-windows 3,364 stars (created 2023-10-14, last 7 days +111), chinese-ime-lm 908 stars (an IME language model the author trained). The whole org totals about 5,500 stars.

- Download volume has evidence; the numbers are GitHub Release asset downloads: android-v0.6.0 about 1,800, linux-v0.14.0/0.15.0 about 1,750–1,860 each, macos-v0.55.0 about 630. iOS goes through TestFlight and isn't on the App Store (searching "水杉输入法" in iTunes returns nothing).

- Development intensity is very high: in the last 65 days the main maintainer houko (Xiao Mo, Tokyo, u-next engineer) made about 22,518 commits. Among contributors there's a `claude` account, and 414 commits carry co-authored-by, clearly driven by an AI coding agent. Releases ship independently for each of the six platforms, with new builds almost daily.

- The founder is FanyFull / fanlusky (Lu Fan, @FlyanLu on X, "苏小凡Channel"), based in Shenzhen. He wrote on V2EX that he was GAP for two years, and "with Coding Agents speeding things up" he built this input method (V2EX t/1232135,2026-08-04).

### II. Competitors

| Product | Form | Launched | Scale | How it differs from Metasequoia |
|---|---|---|---|---|
| Rime (Weasel/Squirrel/Tongwen) + Rime Ice | Open source | From ¥2012 | weasel 8.3k, squirrel 6.5k, rime-ice 20k stars | A veteran customizable framework you configure yourself; candidate translation needs third-party plugins (daocatt/rime-translate, rime-trans). Metasequoia is a complete, out-of-the-box product |
| Qingjian (qingjian-team/qingjian) | Open source, Rust | 2026-09-07 | 7,628 stars, 652 forks, peak single day +277 (10-10), last 7 days +275 | Nearly the same pitch: Rust, translations beside candidates, local small-model reranking. But Qingjian focuses on macOS first, then Windows; Linux/Android are previews. The story is tighter around "learn languages", it has a trendshift badge, and outlets like Tencent News have covered it. **Same idea, and it gets 6.5× the attention of Metasequoia** |
| NanoMouse | iOS App | 2026-01-13 | 18 ratings on the China App Store | iOS only. According to the Metasequoia author, the two posted candidate translation on Xiaohongshu at about the same time and cross-promoted each other (@FlyanLu 2026-10-31) |
| DDUP input method "learn English as you type" | iOS App | 2026-10-16 | 20 ratings | Commercial app chasing the same pitch |
| Z-Power IME (feiyuaq/zcfsrf) | Open source | 2026-10-02 | 267 stars | "Learn English while typing," a follower |
| WindInput, Cassotis, Moqi Windows, aime, CyIME, lime and others | Open source | 2026 years | 300–1,400 stars each (CyIME 677 stars, single-day +214) | A wave of open-source IMEs built with AI coding over 2026 years; most are single-platform or built on Rime |
| Sogou, Baidu, iFlytek, WeChat Input, Doubao Input | big tech | Already around | Hundreds of millions of users | Big-company tools come with ads and privacy worries. Minimal, open source, no data sales: that's Shuishan's pitch. |

### 3. How users came in (growth tactics)

1. **Inherited the old repo and its users** (confirmed): msime was originally called MSIME-Apple (the official docs say "msime (formerly MSIME-Apple, now a multi-platform repo)"). 2026-09 turned it into the main multi-platform repo, and the old Linux and other repos were archived and folded in (msime-linux is marked "migrated to the msime main repo"). The msime-windows README and the website both point to it, so old users and Windows traffic were redirected there.

2. **Xiaohongshu blogger recommendation** (confirmed, though it was for the Windows version): 2026-08-30 the author posted on V2EX, "The open-source Windows Shuishan IME got a traffic explosion after a Xiaohongshu blogger recommended it". The post links to a note by the blogger "晓白和林亦" titled "An IME that lets you memorize words while typing! Completely free!" (V2EX t/1238218). This was the first time the "vocabulary/candidate translation" selling point went viral. Three days later (9-02) the msime main repo was created.

3. **X repost (confirmed)** : @knowledgefxg ("Knowledge Sharer") 2026-10-05 posted a recommendation, original text: "An open-source privacy input method... shows the English meaning beside candidate words, so you learn English while you type", linking the msime repo directly. 76.1K views, 411 likes, 655 bookmarks (read via Jina from x.com).

4. **知乎 KOL 推荐** (已确认):「AI 编程开发话题下的优秀答主」wangleineo 2026-10-20 发想法,说它「比用过的所有中文输入法都好」,夸了 Rust、多平台,还有作者自训的语言模型,得 28 赞,IP 属地日本。第二天 10-21 正好是 msime 单日最高 +93,时间对得上,但不能确定就是这条带来的。

5. **Author self-promotion in communities** (confirmed): V2EX 2026-08-04 launch post (23 replies); 2026-11-01 piggybacked on a hot thread about reverse-engineering the Sogou IME to promote their own project (1,786 views), and directly asked for stars (V2EX t/1253625). A user on linux.do posted an unprompted recommendation (topic 2914721).

6. **The author's own videos** (confirmed, small reach): on Bilibili, "Shuishan IME feature overview (Windows)" and "The second IME for college students: Shuishan IME"; on YouTube, @suxiaofan's "The second IME for Chinese speakers: Shuishan IME". About 58 views in total.

7. **Wide distribution** (confirmed): website msime.app, AUR (msime-bin), openSUSE OBS, a Homebrew tap, the npm package @msime/web-engine, the crates.io crate msime-engine, Pgyer (Android), and TestFlight (iOS). Download sites (32r.com and others) also index it passively.

8. No evidence found of paid acquisition, a Product Hunt launch, Hacker News, the Ruan Yifeng weekly, or HelloGitHub features (the only HelloGitHub hit was a different Rust IME on a tab page).

### IV. Why it took off

- **The selling point hits a concrete scenario** : "English shown beside candidate words" puts vocabulary study inside the typing action people do all day. Xiaohongshu's learner audience and X's productivity accounts are willing to share it. The first line of every external recommendation post leads with this point.

- **There was existing momentum to ride** : the Windows version had been around for 2023 years, and August's Xiaohongshu post had just gone viral. The msime main repo was an "all-platform" release, so from day one it had old users, cross-linked READMEs, and traffic from the website.

- **Open source, minimal, privacy: aimed at the pain points of big-company IMEs** : the README includes a detailed item-by-item list of "what data leaves the device". The author has publicly talked about turning down investors who wanted to sell user data (V2EX 8-30). The landing point of the Zhihu recommendation was also "making it out of passion" and "big-company IMEs are so bad".

- **Native on six platforms plus HarmonyOS** : among open-source Chinese IMEs, it is nearly the only one that ships native installers for Android, iOS, macOS, Linux, Windows, and HarmonyOS at the same time. Rime-based projects are separate frontends per platform and require manual setup; Qingjian is currently mainly desktop.

- **Release speed from AI coding agents** : over 20,000 commits in two months, releases every day, fast issue responses. That alone attracts a developer audience, and the README looks professional (badges, architecture docs, CI).

### Five: What sets it apart, and whether that explains the growth

- The differentiator users notice first (candidate translation / vocabulary study) **is why it grew, but it isn't unique to it** : Nanomouse, Qingjian, Rime plugins, Z-power, and DDUP showed up around the same time or right after. The author also admits it was "the same period".

- What really sets it apart is **one Rust engine plus a native host on each of the six platforms** , along with **a privacy statement you can check line by line** and **an input-method language model the author trained themselves** (chinese-ime-lm, 908 stars). These help it keep developers and Linux/HarmonyOS users around. But they aren't the spreading hook, and they don't explain the daily growth after October.

- Comparing with Qingjian shows that for the same idea, the project with the tighter story ("learn a foreign word while you type") and that polished one platform first got far more stars. Shuishan's story is diffuse: six platforms, the engine, privacy, AI, vocabulary, skins, and plugins are all pitched at once.

### VI. Visibility

It's spreading within niche circles. Chinese developers and productivity communities (V2EX, linux.do, Zhihu, the Chinese X community) have several posts, and the biggest X post got 76K views. No English media coverage, no Trending placement, and no public revenue (it's free, with only a plan for a 16-yuan one-time voluntary support fee not yet implemented, see msime.app/price, plus Afdian). But the "candidate-translation IME" niche is already crowded: Qingjian has 7.6k stars, at least two iOS options (Nanomouse and DDUP), and several Rime plugins.

### 7. What to borrow

- Building a learning feature on top of a high-frequency action (typing) and creating a selling point that one screenshot can explain suits spread by productivity accounts on Xiaohongshu and X.

- When an old product expands to new platforms, creating a fresh "all-platform" main repo and cross-linking the old READMEs and website to it turns existing stars into new ones.

- Turning privacy into an auditable document (what leaves the device, defaults, how to turn it off) works as trust-building in a category where big-company products have poor reputations.

- One person with an AI coding agent can ship a native six-platform product, with the technical barrier greatly lowered. This is also why so many similar projects appear at once.

### 8. What's still uncertain

- The specific source of 10-21's +93 single-day gain is unconfirmed. The Zhihu post (10-20) matches the timing, but it had only 28 likes, and there may be other channels, such as new Xiaohongshu notes, which we can't access.

- We can't access the specific number of Xiaohongshu notes or their engagement. The V2EX post says "traffic exploded" but gives no numbers.

- Daily star curves from the early period (9-02 to 10-17) aren't available, so the lift from 10-05 X posts can't be quantified.

- Actual daily active users: the README says all six platforms send anonymous daily-active statistics to api.msime.app, but no numbers are published.

- Why Qingjian grew so much faster (single-day +277): not investigated in depth this time; it may be related to trendshift and Tencent News coverage.

### Sources

- Data card: /home/ljr/fanout/why/cards/gh_metasequoiaime__msime.json

- GitHub repo_profile (2026-11-06 queries): msime, msime-windows, chinese-ime-lm, qingjian, zcfsrf, and others

- GitHub API: repos/metasequoiaime/*, contributors, releases, qingjian-team/qingjian

- V2EX: t/1232135 (2026-08-04 launch), t/1238218 (2026-08-30 Xiaohongshu blew up), t/1236492 (2026-08-22 user shares the vocabulary feature), t/1253625 (2026-11-01 self-promotion)

- X: [x.com/knowledgefxg/status/2107071736533672391](https://x.com/knowledgefxg/status/2107071736533672391) (2026-10-05,76.1K views); [x.com/FlyanLu/status/2116621416053563516](https://x.com/FlyanLu/status/2116621416053563516) (2026-10-31, Nanomouse)

- Zhihu: [www.zhihu.com/pin/2095836204663157255](https://www.zhihu.com/pin/2095836204663157255) (2026-10-20)

- linux.do:[linux.do/t/topic/2914721](https://linux.do/t/topic/2914721)

- Website: [msime.app/](https://msime.app/) and [msime.app/price/](https://msime.app/price/)

- Qingjian: [qingjian.app/](https://qingjian.app/) and Tencent News [news.qq.com/rain/a/20261006A06DIO00](https://news.qq.com/rain/a/20261006A06DIO00)

- App Store China search: Nanomouse NanoMouse (id 6757662900), DDUP IME (id 6819322032)

This article was written by AI from public data and the data we collect daily. Numbers are sourced, and inferences are labeled.