Stringers
jizizr/mhub · jizizr

Listen to Apple Music lossless without logging in: a web player on one free instance hits 170 stars on linux.do in a week

What it is A web music player you host on your own server. It merges Apple Music, NetEase Cloud Music and QQ Music into one queue, copies the Apple Music interface, and the author runs a public URL that needs no login.

Why it blew up The author posted in linux.do's "open-source promotion" section and shared a public URL that lets people listen to lossless Apple Music in the browser with no login and no subscription. The freeloader-friendly community thanked him with stars, and a Chinese X repost account with 136k followers spread it to a second wave 11-12.

CHow you could follow Don't build a copycat music player. Learn the playbook instead: open source + a self-hosted public instance with no login + posting in the right community. Apply it to a compliant topic, or split the browser ALAC lossless part into a standalone library.

C grade: the playbook is worth learning, but its stars came from "free lossless official music", and copying it means putting up your own account and storage and betting it won't get taken down. The category leaders have tens of thousands of stars, and several projects in the same community launched the same week. Whispers in niche circles
Data (updated daily) · Repo created 2026-10-22 · 212 stars · 7 days +88
GitHub · stars
5010015020011-0911-1211-1511-1711-20212
Data 2026-11-09 → 2026-11-20 (12 days, 12 observations). Points before 10-28 come from deeply / archive backfill, not every day has one

A young developer, not well known on GitHub, copied the Apple Music web interface, hooked it up to his own Apple Music account, and put up a URL anyone can open. Visitors can listen to full ALAC lossless tracks. A week after posting on linux.do, the repo went from dozens of stars to 179. Below: what it is, where the stars came from, why copying it isn't worth it, and which parts of the playbook you can take.

A Saturday-night post

On 2026 Nov 7 23:52 (Beijing time), linux.do user zrcc posted in the "Dev & Tuning" section with the title "Mhub, a free and open-source online music platform," tagged "Open-source promotion." The first paragraph gave two links: the GitHub repo jizizr/mhub and the live site music.op.wiki. Right after came one line: "Guests and logged-in users get exactly the same songs and audio quality." To demo the share feature, he pasted Stefanie Sun's "我怀念的" ("What I Miss").

One minute later came the first reply: "Support, bro." The next morning, around 6 a.m., someone asked "Can I download it?" 16 minutes later the author replied: added a download button, batch download has an API doc, "you can have your AI write a script from the doc." By 11-08 in the evening, the replies under the post had turned into this:

"Beautiful front end, and you can listen to Apple Music for free" (floor 8) "QQ Music has been painful; without money I can't listen to a full song" (floor 9) "This is the absolute best part, no login wall" (floor 30, quoting the author's line that guests and logged-in users are no different) "Good stuff, starred and deployed right away" (floor 26)

As of 11-16, the thread had 85 floors, 62 participants, 203 likes and 1,847 views (linux.do, 11-16).

What it is

MHub is a web music player you host on your own server. Created in 10-22, AGPL-licensed, with a Rust backend and a Vue 3 frontend. It merges search and playlists from Apple Music, NetEase Cloud Music and QQ Music into one playback queue. A NetEase playlist can be played with Apple Music audio, and after each track it picks the next one following Apple Music's radio logic. The interface copies Apple Music's web version, with word-by-word lyrics, cover-based color backgrounds, and sharing that exports lyrics as posters.

The hard part is the audio source. Apple Music lossless needs a decryption service the author built himself, plus a real Apple Music account. The idea comes from download tools like apple-music-downloader. Decrypted full tracks are uploaded to Telegram for storage, so the next person who clicks the same song gets it directly. Browsers generally can't play ALAC, so it compiles a Rust decoder to WebAssembly and decodes in the page. The author said in the thread that each song takes about 100MB on the backend (floor 51). The README states: "This repo provides no accounts, sessions or media files."

The author jizizr joined GitHub in 2021, has 42 followers, and his public repos include small tools like a Chongqing University of Posts and Telecommunications timetable. He looks like a student or recent graduate (inference). His previous most-starred repo had only 21 stars; MHub is the first to break out. 11-09 He has since opened two more repos for Android and iOS clients.

This track was already crowded

"Aggregate multiple platforms, listen to music for free" is one of the oldest topics in China's open-source scene. The leading projects all have tens of thousands of stars:

ProjectLaunchedstar(11-16)How MHub differs
lx-music-desktop (Lo Xue Music)2019-0855kDesktop client; users must import their own source scripts
YesPlayMusic2020-1034kGood-looking, but only connects to NetEase Cloud Music
MusicFree2022-0829kPlugin-based; the core doesn't ship with any source
AlgerMusicPlayer2023-1018kNetEase desktop player, focused on downloads
listen12016-0412kLong-running Chrome aggregator extension
go-music-dl2026-016,400This year's hit, focused on downloading from across the web
navidrome201625kProper self-hosted; only plays your own music files
MeloX and others2026-07/08A few hundredAlso promoted on linux.do, also imitates the Apple Music interface, but has no AM lossless

The same week on linux.do, there were also "MyMusic multi-platform aggregation", a player with a "joined membership pool" note, and a post about "wanting to build a community charity music app", all competing for the same crowd.

MHub didn't invent a new category. It filled a combination that was open: official Apple Music lossless audio + the original Apple Music interface + listen right in the browser, with no client to install and no sources to configure. Other projects hand you a shell and leave you to find audio yourself; MHub prepares both the audio and the server for you.

Where the stars came from

GitHub daily stars:

DateWhat happenedstar
11-07 late nightThe main linux.do thread goes live(earlier numbers unavailable)
11-08The thread's hottest day; most replies came this day—
11-09The author shares the iOS client in the thread63
11-10Posts another thread, "The Mhub client is here" (257 views)79
11-11101
Night 11-12A Chinese X account with 136k followers, "Namis Knowledge Base", recommends it: 11,836 views, 123 bookmarks107
11-13124
11-14155 (+31, the single-day peak)
11-15Another account with 48.5k followers nearly copy-pastes that tweet and reposts it, 1,352 views159
11-16172; GitHub live 179 stars, 34 forks

Both waves are clear. The first was the linux.do post, which brought in the first 60+ stars. linux.do's open-source promo section has rules: declare the project is fully open source, link the community that endorsed it, and post screenshots with the AI-polished parts flagged. The people who read that section deploy things, understand lossless audio, and love free stuff. They're the same crowd as MHub's users. The second wave was Chinese repost accounts on X. These accounts specialize in scanning linux.do for "free open-source tools" topics. The author didn't bring them in. The timing of 11-13 and 11-14 matches their pickup. A few aggregator sites also reposted the thread, and a community site wrote an article titled "100MB single-track caching makes MHub's self-hosting cost fall on storage." All of that is long tail. I found no traces on HN, Reddit, V2EX, or Bilibili, and it never made GitHub trending.

34 forks against 179 stars is close to one fork for every five stars, which is very high for a new repo. Most fork accounts have Chinese names, which lines up with the thread's string of "deployed it" replies. Many people weren't there to read the code. They wanted their own copy running.

Users came for the result, not the architecture

The README has full architecture and data-flow diagrams: chunked decryption, playback starting after the first chunk, Telegram as a free file store, two <audio> elements that take turns preloading so playback doesn't cut out when the phone locks, and Yjs for playlist sync. The engineering is clever.

But in the 85 replies, almost nobody talks about these. People ask "where does the audio come from" (floor 19), "can I download it", "is there Spotify", "can I play it in the car", "is this audio quality really hi-res". The worries are the same: "is there a copyright issue" (floor 27), "feels like a lawyer's letter is coming" (floor 29), "just don't let it spread too far or it'll get taken down" (floor 49), "hope it lasts a few more years, a lot of the free music sites I bookmarked have already shut down" (floor 66).

So these stars read more like users saying "thanks" for a free service than developers voting for a project. The author replied fast: downloads added that night, a client out by day three, and an immediate fix for Safari playback. That kept the "thanks" going for more than a week.

Its weak spots, which are also traps for anyone who follows it

  • The whole thing runs on one person's accounts and servers. Each 100MB track's storage and bandwidth come out of the author's pocket. Apple Music is wired to a China-storefront account, so songs without rights in the China store won't play (floor 56). A banned account, a broken decryption service, or a platform tightening its anti-abuse checks could each take the public instance down.
  • The legal risk is spelled out in the comments. Decrypting lossless audio from a paid platform and publishing it with no login sits in a gray zone. The repo could get a takedown notice at any time, and the stars would go with it.
  • Compatibility is still rough. Safari throws "the browser cannot decode this audio stream" (floor 70). The iOS client originally required system 26 (floor 74; later lowered).
  • Others are already building on it. On floor 79 someone "vibe-coded an Android version on top of the author's work." On floor 75 someone wants to "pool a few friends' memberships and share them." If you build a clone now, others are already standing on its shoulders.

If it were you

Verdict: C. You can learn a very effective playbook for getting stars from this, but cloning it isn't worth it. Its stars came from "free lossless Apple Music," and to get the same result you'd have to pay for an account yourself, carry storage and bandwidth, and bet you won't get taken down. The category's leaders are established projects with tens of thousands of stars, and in the same week three or four people in the same community were building the same thing. 179 stars is respectable, but it's far from those leaders, and no one can say whether it survives the next takedown or ban.

Who it suits: Students or young developers who want their first breakout open-source project, who are well connected in circle communities like linux.do, who have a cheap server and are willing to run a public instance, and who can reply to comments every day for a week and ship changes the same day.

Who it doesn't suit: Anyone building the tenth "multi-platform aggregator music player." The idea is worn out, and the only differentiator left is the audio source, which carries the biggest legal risk. Anyone afraid of lawyer letters, or unwilling to pay server costs for free users. Anyone who thinks a good README and architecture diagram will earn stars. Nobody in this case looked at the architecture.

How to do it (ranked by how worth it each is):

  1. Take the playbook, change to a compliant idea. "Open source + the author runs a public instance with no login + post in the right community" works for anything other people pay for or log in to access, that you can offer for free. The condition is that the "free" part can't depend on cracking someone else's paid content. Getting users to try it first and then starring is much easier than asking them to "self-host it yourself."
  2. Extract its engineering parts and ship them standalone. Few good standalone libraries exist for playing ALAC lossless in the browser right now. The old alac.js has 213 stars and was last updated 2014 years ago. Another ALAC decoder, written in Rust and compiled to WASM, was created this past September and has 0 stars (GitHub, 11-16). Packaging "lossless ALAC playback in the browser + no interruption when the screen locks" as an npm package with a live demo page carries no copyright risk, and the audience is everyone building web music players. The other two ideas (Telegram Bot API as a free large-file store, streaming decryption) could each become a small library too. The ceiling is modest, probably a few hundred stars (estimate).
  3. Not recommended: building a direct MHub clone. If you do it, aim only at audiences it doesn't cover, like the Spotify sync or car-head-unit requests people raised in the thread. Those demands come from the same "free-rider" users and carry the same risk.

How to get users: Follow its route. Post in linux.do's open-source promo section and follow the section rules (declare it's open source, flag AI-polished parts in screenshots). Make the first post link to a live address that opens right away, and pick the best screenshots. Watch the comments for the first two days, and build any request people raise the same day. On days three and four, post again with something new (a client, a new feature) to keep the momentum going. Chinese repost accounts on X will find you on their own. You don't need to chase them.

When is "before" the deadline: The playbook has no time limit. The standalone-engineering option is best done while people are still discussing MHub this week. Ship it before the end of 2026, and mention it in the MHub thread or its issues to catch the same people (estimate).

What would show this read was wrong: ① If MHub climbs to several thousand stars within a month or two and its public instance stays up with no takedown notice, the legal risk and star ceiling in this category are looser than estimated here, and a clone would be worth building. ② Conversely, if the public instance shuts down or the repo gets removed, C should drop, and even the "follow the playbook" route is off the table. ③ If someone releases "browser ALAC playback" as a standalone library first and gets a few hundred stars, the path in option 2 is taken.

Data and sources

GitHub: the jizizr/mhub repo page, README, and author profile. Daily stars (11-09 to 11-16) and live counts (179 stars, 34 forks, 11-16). Star counts for comparable repos (11-16). linux.do: the main thread topic/3138232 (85 replies, 203 likes, 1,847 views, 11-16) and the client thread topic/3147924. Floor quotes all come from the main thread. X: the "Namiknowledge" account @fhwofjow51260 11-12 post (11,836 views, 107 likes, 123 bookmarks) and @0x8888 11-15 post (1,352 views). Aggregators and long tail: aicodeok.com, RepoCloud, trendshift. Star counts and the day-by-day curve before 11-07 aren't available, so how much each wave contributed is inferred from daily additions. Wherever this says "estimate," it's an inference.

GitHub repo ↗

Research notes: competitor table, acquisition evidence, full research report
Competitors
NameWhereLaunchedScaleHow it differs from this one
lyswhut/lx-music-desktop (Lxmusic)Open source2019-0855.4K starDesktop client with custom audio-source scripts, no audio included. MHub is a web app plus public instance, with AM lossless.
qier222/YesPlayMusicOpen source2020-1033.6K starA high-polish desktop client for NetEase Cloud Music only. MHub is multi-platform and pulls audio from Apple Music.
maotoumao/MusicFreeOpen source2022-0828.6K starPlugin-based. The core has no audio source. MHub ships its own audio and an online instance.
algerkong/AlgerMusicPlayerOpen source2023-1017.6K starDesktop NetEase Cloud Music player, focused on downloads and desktop lyrics.
listen1/listen1_chrome_extensionChrome2016-0412.1K starLong-established multi-platform aggregator extension. No AM lossless, can't self-host.
guohuiyuan/go-music-dlOpen source2026-016.4K starWhole-web search-and-download tool (10+ domestic platforms), focused on downloading. MHub focuses on online playback and an Apple-style experience.
navidrome/navidrome / jeffvli/feishinOpen source2016 / 202224.8K / 10.4K starOrthodox self-hosted server that plays only your own library. MHub pulls audio from streaming platforms.
MeloX-Android / youshen2/MeloXOpen source2026-07/08337 / 425 starAn Android client with an Apple Music-like interface, also promoted on linux.do. No AM lossless, no web instance.
zhaarey/apple-music-downloader / glomatico/gamdlOpen source2024 / 20222.4K / 2.8K starA CLI that downloads AM ALAC. MHub's wrapper decryption idea comes from this family. They have no player.
Apple Music web / CiderBig-company features—Official ¥12/month. Cider has 1 7.1K stars.The official service requires a subscription. MHub copies the Apple Music interface, and its public instance is free.
Where users come from
  • Verified linux.do open-source promo post (author zrcc posted it on their own): On 11-07 they posted "Mhub, a free open-source online music platform," linking to music.op.wiki. 85 replies, 62 participants, 203 likes, 1,847 views. Comments were mostly "already starred," "deployed it," and "free Apple Music listening."
  • Verified linux.do second thread sustaining momentum: 11-10 "Mhub clients are here," launching Android and iOS apps, 257 views.
  • Verified X Chinese repost accounts: 11-12 Namiknowledge (136k followers) recommendation post, 11,836 views, 107 likes, 123 bookmarks, matching daily gains of +17/+31 on 11-13/14. 11-15 Cevenif (48.5k followers) copied the copy and got 1,352 views.
  • Verified Aggregator and SEO long tail: repost articles on feedbbs, ifeed, and locdd; an aicodeok review article; a RepoCloud one-click deploy page; listings on repoalerts and trendshift.
  • Inferred GitHub trending: trending_days_any=0, did not chart.

Why jizizr/mhub (MHub) took off

Research date 2026-11-16 (the brief says 11-07, data card refreshed to 11-16). Channel: GitHub.

Bottom line

MHub nominally calls itself a "self-hosted music player," but its stars didn't come from self-hosting. In the linux.do post, the author linked a public instance they run, music.op.wiki. No login and no membership needed. You can listen to Apple Music ALAC lossless and NetEase Cloud Music in the browser, with an interface copied from the Apple Music web app. Users on linux.do who love free stuff saw that and showed up. A Chinese repost account on X with 136k followers shared it once and brought in a second wave. The stars read more like users saying "thanks" for a free service than developers voting on the architecture.

What it is

  • Repo: github.com/jizizr/mhub, created 2026-10-22, AGPL-3.0, Rust (Axum/Tokio) backend with a Vue 3 frontend. 11-13 total commits as of 73 (GitHub API).
  • Features (from the README): search, playlist import, and cross-platform matching for Apple Music, NetEase Cloud Music, and QQ Music (the thread also mentions YouTube Music), all feeding one queue. Continuous playback uses Apple Music's AutoPlay radio. ALAC is decoded in the browser by a Rust decoder compiled to WASM. Two <audio> elements preload so playback continues in the background after the screen locks. Word-by-word lyrics borrow from AMLL. Audio and lyrics are cached in IndexedDB, and the app installs as a PWA. Sign-in uses passkeys, and playlists sync in real time via Yjs.
  • Key upstream dependency: Apple Music audio relies on a self-hosted wrapper/wrapper-lite decryption service (the approach comes from the zhaarey/wrapper and apple-music-downloader family of tools), plus a real Apple Music account. The decrypted full M4A is uploaded to the Telegram Bot API for storage, with the file_id saved in MySQL so the next request is a direct hit. The README states: "This repo does not provide any accounts, sessions, or media files."
  • Author: GitHub user jizizr (Allen Joe), account created 2021, 42 followers, 67 public repos, blog zr.wiki. Other small projects include a Chongqing University of Posts and Telecommunications timetable tool and a HyperOS pickup-code tool. Looks like a student or young developer. The data card says "owner_repos 1", which is wrong. The author's previous top repo, signaldock, had only 21 stars, so MHub is the first project of theirs to break out. 11-09 They have since opened two clients, mhub-android (7 stars) and mhub-ios (2 stars).

Growth data and timeline

DateEventstars (our GitHub repo_day)
10-22Repo created—
11-07 23:52 (Beijing time)The author, zrcc, posted "Mhub, a free and open-source online music platform" on linux.do in the "Dev Tuning" section, tagged "Open-source promo", with a live instance at music.op.wiki—
11-08The most active day of the thread (floors 4–60 are mostly from this day): "starred already", "deployed it", "free Apple Music listening"—
11-0963
11-10The author followed up with "Mhub client is here" (Android/iOS, floor 7, 257 views)79(+16)
11-11101(+22)
11-12 20:20 (Beijing time)On X, "Nami's Knowledge Base" (@fhwofjow, 51.3K followers) posted a recommendation: 11,836 views, 107 likes, 123 bookmarks, 18 reposts (fxtwitter API)107(+6)
11-13124(+17)
11-14155 (+31, single-day high)
11-15On X, "Cevenif" (@0x889M followers) reposted nearly word-for-word the same copy: 1,352 views, 9 likes159(+4)
11-16172 (+13); live GitHub API: 179 stars / 34 forks
  • Current numbers for the linux.do main thread: 85 replies, 62 participants, 203 likes (173 on the opening post), 1,847 views (linux.do topic JSON, scraped 11-16).
  • +109 stars in the last 7 days (data card). Fork/star ratio is about 19%, high for a new repo. Most fork accounts have Chinese usernames, which fits the "community members deploying their own copy" story.
  • It has never made GitHub trending (data card trending_days_any=0), but trendshift.io already indexes it on similar pages.
  • We can't get the early star curve. The REST stargazers endpoint returns 404 for this repo, GraphQL stargazers.totalCount returns 0, while stargazerCount is 179. The same author's other repos work fine. We haven't confirmed the cause of this mismatch (possibly GitHub restricting or caching the stargazer list for this repo). So we can't verify how many stars it had before the post. We can only infer that most came after 11-07 and 11-07.

Where users come from (growth levers)

  1. linux.do open-source promo post (confirmed, main channel). linux.do has a compliant "Open-source promo" section (requires a full open-source declaration, linking to community-recognized sources, and screenshots if AI polishing was used). Its users are exactly the domestic developers who love free stuff and self-hosting. The replies show the motivation clearly:
    • "The front end is gorgeous, and you can listen to Apple Music for free" (floor 8)
    • "QQ Music has been a pain, you can't listen to a full song without paying" (floor 9)
    • "Guests and logged-in users get the same songs and the same audio quality. That's the most amazing part, no login wall" (floor 30)
    • "Starred it and deployed it right away" (floor 26), "Deployed it on my server" (floor 4)
    • Worried voices: "Could there be copyright issues?" (floor 27), "Feels like a lawyer's letter is coming" (floor 29), "It just can't spread too widely or it'll get taken down" (floor 49)
  2. X Chinese repost accounts (confirmed, second wave). 11-12 The Namida knowledge-base tweet lines up in time with the spike in 11-13/11-14 (+17, +31). 11-15 Cevenif copied the same copy almost word for word, so its impact is small. Both accounts are reposters pushing "free open-source tools" type content, not the creators' own accounts.
  3. Aggregator sites and SEO reposts (confirmed, small volume). feedbbs, ifeed, and locdd scraped the linux.do post. aicodeok.com (AIDE community) wrote an article, "100MB single-track cache puts MHub's self-hosting cost on storage". repoalerts, the RepoCloud one-click deploy page, and trendshift also index it. All of this is passive long tail.
  4. No hits on HN, Reddit, V2EX big posts, Bilibili videos or press coverage (Jina/Exa search). Our X database (x-radar) doesn't turn up mhub either, so the corpus may not cover these two accounts.

Competitors

CompetitorsWhereLaunchedScaleHow MHub differs
lyswhut/lx-music-desktop (Lxmusic)Open-source desktop/Android2019-0855.4K starCustom source scripts plus a desktop client, no built-in source. MHub is a web app with public instances and lossless Apple Music.
qier222/YesPlayMusicOpen-source desktop/web2020-1033.6K starA stylish third-party NetEase Cloud Music client that only supports NetEase. MHub is multi-platform and has an AM source.
maotoumao/MusicFreeOpen-source Android/desktop2022-0828.6K starPlugin-based, the core has no source built in. MHub ships its own source and public instances.
Binaryify/NeteaseCloudMusicApiOpen-source API2016-0630.2K stars (archived)NetEase Cloud Music API layer, which many players depend on
algerkong/AlgerMusicPlayerOpen-source desktop2023-1017.6K starA third-party NetEase Cloud Music player with downloads and desktop lyrics
listen1/listen1_chrome_extensionChrome extension2016-0412.1K starAn established multi-platform aggregator extension. MHub is a self-hosted server plus PWA.
guohuiyuan/go-music-dlOpen-source CLI/web2026-016.4K starThis year's hit: full-web music search and download (10+ domestic platforms), focused on downloads. MHub focuses on online playback and the AM experience.
navidrome/navidrome, jeffvli/feishinOpen source, self-hosted2016 / 202224.8K / 10.4K starProper self-hosting, only plays music files you own. MHub's "self-hosting" actually proxies streaming platforms.
SPlayer-Dev/SPlayerOpen-source desktop2023-017.4K stars (archived)NetEase Cloud Music plus Jellyfin/Navidrome
lladlam/MeloX-Android, youshen2/MeloX, Synoc-MusicOpen-source Android2026-07/08337 / 425 / 1 starAlso promoted on linux.do, also an "Apple Music-like third-party music app" with the same UI approach, but no lossless AM source.
zhaarey/apple-music-downloader, glomatico/gamdl, WorldObservationLog/AppleMusicDecryptOpen source CLI2024 / 2022 / 20242.4K / 2.8K / 717 starApple Music ALAC download tools. MHub's wrapper decryption idea comes from this lineage. They only download and have no player UI.
Cider(ciderapp)Desktop (Cider 2, paid)2021Cider 1 has 7.1K starsA third-party client for official Apple Music, requires a subscription
Apple Music web version, music.apple.combig tech——MHub copies its interface. The official service costs ¥12/month.
Contemporaneous linux.do posts: "MyMusic multi-platform aggregator for Android" (3141836), "Music Player v1.3.5: now with a member account pool" (3134146), "I want to build a community-exclusive charity music app" (3133547)Community projects2026-11Small scaleIn the same community, several "free member songs" projects competed for attention in the same week. MHub's edge is lossless AM plus an Apple-style interface.

Summary: the "multi-platform aggregation, free listening" category is already crowded, with leaders at tens of thousands of stars. MHub didn't enter with a new category. It had a combination nobody had claimed: official lossless Apple Music audio, the Apple Music interface as-is, and playback right in the browser (no client install, no source setup).

Why it took off (causes)

  1. Public instances let people get lossless Apple Music for free, no setup. Most domestic aggregator players rely on NetEase, QQ, or Kugou APIs, so members-only songs only offer preview clips, and users usually have to import their own sources. MHub's author used his own China-region AM account with a wrapper to play AM ALAC directly in the browser, and guests can listen too (floor 51: "I checked my backend storage, one song takes 100MB"; floor 56: "I hooked up a China-region Apple Music account"). The most common comment was "free Apple Music listening".
  2. Right community. linux.do's "Open-source promo" section exists to drive traffic to open-source projects, and its users match the product exactly (they self-host, love free stuff, and care about lossless audio). The main post got 173 likes and 1,847 views, which brought at least the first 60+ stars.
  3. Polish and design passed the bar. The interface is a near-exact replica of the Apple Music web app, with word-by-word lyrics, cover-color backgrounds, and lyric poster sharing. The README has a full architecture diagram in mermaid. Comments like "gorgeous front end", "very complete", and "love the interface" were common, showing users cared about the experience, not the tech.
  4. The author responded fast and iterated fast. When people asked for downloads, a download button shipped the same day (floor 7). The Safari bug was fixed on the spot. Within three days, Android and iOS clients followed with a second post, which kept the momentum going.
  5. An X repost account amplified it. An account with 136K followers brought roughly 12K views and 123 bookmarks, matching the spikes on 11-13 and 11-14. These repost accounts actively hunt for "free and open-source" topics on linux.do. The author didn't buy them.

Differentiation, and whether it's a manufactured ranking

  • Differentiation: ① Source layer pulls Apple Music ALAC lossless. Decryption starts playback as soon as the first segment is decrypted, decoupled from the caching path, with Telegram serving as a free file store. ② Decodes ALAC via WASM in the browser, so the web player delivers true lossless (most other web players only handle AAC/MP3/FLAC). ③ UI is a pixel-level copy of Apple Music. ④ Public instance funded by the author out of pocket. Guests and logged-in users get exactly the same audio quality.
  • Is that the reason: Yes, but what drove it was ① plus ④ delivering the result to users, namely "free, genuine lossless listening". Not the engineering stack itself (Rust, WASM, Telegram caching). The comments barely discuss architecture. People ask "where does the source come from", "can I download", and "will I get sued". The engineering makes the thing work and stay stable, but users came for the outcome.

How many people know about it

  • It spread within a small circle: one linux.do post with 1.8K views, two Chinese X repost tweets (12K and 1.35K views), and a few aggregator sites. It never made GitHub trending, got no media coverage, and didn't reach HN or Reddit. It's free and open-source, so there's no revenue.
  • Copycats: no copycat repo spotted so far. The 34 forks are mostly for personal self-hosting. Cevenif reposting Nami's tweet copy is just secondary spreading. MeloX, MyMusic, and the "member account pool" player from the same community are parallel competitors, not copies of MHub.

What others can learn from it

  • "Open source plus a public instance the author runs" is a strong GitHub star strategy. Users get to use it for free first, then come back to star it as thanks. That converts far better than "please deploy it yourself". The cost is that the author carries server, storage (100MB per song), and legal risk.
  • In a tightly defined community like linux.do, post an "Open-source promo" thread that follows the section rules. Spend the first two days watching replies and filling in features, then renew momentum with a second post (here, the client).
  • Some engineering points are reusable: decoupling chunk decryption from caching so playback starts after the first chunk; using the Telegram Bot API as a free large-file store; double <audio> preloading to fix mobile lock-screen playback drops; and Yjs for playlist sync.

Uncertain points

  • Star counts and the daily curve before 11-07 are unavailable (the stargazers API is misbehaving), so how much came from linux.do vs. X can only be inferred from daily gains.
  • The traffic of the public instance music.op.wiki is unknown; it isn't in Tranco's top 1M, so our traffic database can't find it.
  • The upstream could be cut off at any time: Apple could ban the account, the wrapper could break, NetEase could tighten risk controls, or a DMCA takedown or lawyer's letter could arrive. Commenters have already raised this warning. How far the project can go depends mostly on this.
  • The x-radar database didn't capture these two tweets; we haven't verified whether there are other reposts on X.

Sources

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