MyGo Cloud Music: a 8MB NetEase Cloud Music client that lived for 10 days; stars grew faster after a cease-and-desist takedown

In 2026 Oct 26, EGOIST, the developer who wrote tsup, posted a video on X of a Go-based NetEase Cloud Music desktop client: "download size only 8MB, background playback memory only 100MB." The post got 88.8k views. On Nov 4, he posted "got reported 😂" with a screenshot of NetEase Cloud Music's letter, which got 55.4k views. Over those 10 days the repo picked up over 600 stars. Two days after it went offline, nearly 200 more stars came in, and forks rose from 96 to 369. This piece covers why it took off, why it was doomed to go offline, and if you want a project like this that collects stars, which parts you can copy and which you can't.

A very persuasive takedown notice

Open github.com/egoist/mygo-cloudmusic today and the code is gone. All that's left is a commit called "offline" and a README, and the repo is archived. The README's first line is "Native, quietly offline." Then:

NetEase Cloud Music sent a letter requesting that this repo "go offline immediately." The upside of writing it in Go: it starts fast, and it goes offline even faster.

The feature list is still there, just with every word rewritten: "Daily Recommendations: updated every 6:00 no more. Today's Recommendations: official client" and "Heartbeat Mode: turns on automatically the moment the letter arrives." The install section says:

Please use the official client. It's Chromium-based, about 354 MB installed, and uses 700 MB of memory when opened, but it's licensed.

The build section reads: "Nothing left to build. Build time is cut to 0 seconds, the biggest one-time performance win in this project's history." Last line: "Thanks for these 684 stars."

The README itself has been passed around. Users nicknamed this ending the "standard ending," the familiar fate of third-party NetEase clients.

What it is

A third-party NetEase Cloud Music desktop client that runs on Mac, Windows, and Linux. Log in with your own NetEase account by scanning a QR code and play whatever your membership allows: five audio quality tiers up to Hi-Res, daily recommendations, personal FM, heartbeat mode, word-by-word lyrics, a mini player in the notch, and local speech recognition to subtitle podcasts. One user called it "exactly the third-party NetEase client I've been wanting."

The biggest difference from its peers is under the hood: it's not a web wrapper. Most peers are Electron plus Vue, which is essentially a browser. It uses MyGo, a Go desktop framework EGOIST had just released in Sep 26, with the UI drawn directly on the GPU and no WebView. The result is a 8MB program.

Why this one took off

The author brought his own traffic. EGOIST is a long-standing open-source author whose projects total about 82k stars, and his Chinese X account @localhost_4173 has 16.7k followers. On the day after the launch post, 10-28 gained 174 stars in a single day, the highest single day of the whole period.

A set of numbers anyone can read. 8MB vs 354MB, 100MB vs 700MB. No need to explain what GPU rendering is; anyone who has been eaten alive by an official client's memory use gets it. The replies under the launch post said it best: "Same song playing, the pro's client uses only 100 MB, while NetEase Cloud Music eats over a gigabyte" and "Right, the official client's memory footprint keeps growing."

High completeness, not a demo. Most of the features listed above took third-party clients several versions to add. After launch the author kept posting feature videos, for example 11-02's "show hot comments during interludes" (4,690 views).

The name is a joke. MyGo refers to the anime "BanG Dream! It's MyGO!!!!!", and the demo videos play "Haruhikage" and "Mayoi Ga Michi." How much this helped spread it is an inference.

What's really being sold is the framework. Cloud Music is the flagship sample app for the MyGo framework and sits first in the Showcase on the MyGo site. In a little over a month the framework collected 5,000 stars (GitHub page, 11-06), and it gained over 1,100 in the last week alone, several times what Cloud Music gained. "Look, you can build this with MyGo": Cloud Music is the proof of that line.

Where users come from

Checking against the daily star counts (UTC):

A few points: the buzz was almost entirely in Chinese developer circles (X, linux.do, studygolang, V2EX). There was no media coverage, and it never appeared on Hacker News or Reddit. GitHub Trending doesn't list repos younger than about 60 days, so it never showed up there either. The nearly 270 extra forks after takedown came after the repo history had been wiped, and most only got a copy of the README, more like a reflex of "save a copy first" than real use.

Who it's competing with

The third-party NetEase Cloud Music / multi-source music desktop client is a crowded, old niche (around 11-06 stars):

Native, cross-platform on three OSes, small size: almost nobody else in this cell. But the outcome of this path is also in the table: the API library is archived, SPlayer is archived, and YesPlayMusic has stopped updating. The author of NeteaseCloudMusicApi left a comment under the launch post: "Didn't expect my NeteaseCloudMusicApi to still be shining two years after being archived."

After the takedown

On Nov 6, the author released a successor, MyGo Music (egoist/mygo-music, AGPL-3.0): same framework, still no WebView, but the core only plays local files, with accounts, catalogs, podcasts, and radio all handed off to plugins. A plugin is a folder plus a JS file. Tag the repo mygo-music-plugin and it appears on the official site's plugin page, and there's a one-click mygo-music://install-plugin install link. This is the same risk-avoiding approach FeelUOwn and Lx Music's "custom sources" took long ago: a clean core, with online sources left to the community to plug in.

Followers came in even before the takedown, all with the same direction: "same framework, different music source." Examples: mygo-qyin (the fork switched to QQ Music, self-reported by its author under the launch post), a linux.do post on a "QQ Music third-party client based on MyGo," and "Mygo Universe 1 - MySpotify desktop client," plus forks that remove NetEase and keep maintaining. All of these have far less buzz than the original.

If it were you

C. There's a lot to learn here, but little you can take away: what pushed it was the author's own reputation and his own framework, neither of which followers have. What it did (a third-party NetEase client) is a path everyone knows the end of. Build a "MyGo-style QQ Music" and you won't get the stars, and you'll get the same letter.

Who it's for: developers building a new framework, new runtime, or new UI library who are stuck on how to make people see what's good about it. For you, this is a playbook.

Who it's not for: anyone who wants to just build a third-party music client (crowded, doomed to draw legal letters, and most forks after takedown are empty shells); and anyone with no traffic whose posts nobody sees (Cloud Music's peak followed the author's launch post, and without that launch step it wouldn't have taken off).

How to do it: copy the structure, not the topic. Give your framework a mainstream sample app: pick something everyone uses that is notoriously bloated, build a lightweight stand-in, and compress the gap into a set of numbers (8MB vs. 354MB). Put those numbers on the first screen of the README and in the launch video. Choose topics without copyright landmines where you can: your own local files, open protocols, services with public APIs. Or build a plugin architecture from day one like MyGo Music, so the core only ships the legal parts.

How to get users: launch on your own X with a demo video. Chinese developer communities like linux.do, V2EX, and studygolang will repost it on their own. After launch, post a feature video every few days to keep the heat going. Get the sample app into the first slot of the framework's official Showcase so the app and framework feed each other traffic.

Deadline: there's no fixed deadline. But the spot of "a new framework's first killer sample app" only has value in the month or two after the framework launches, and this round is already taken by the author himself.

Where this call would be wrong: ① MyGo Music's plugin ecosystem grows, and some third-party plugin (say, a music-source plugin) gets thousands of stars on its own. That would mean the "do the part it did badly" path is open, and the rating should go up. ② An unknown developer uses the same playbook (lightweight stand-in plus a numbers comparison) and gets thousands of stars within 30 days. That would mean the author's reputation wasn't the whole story.

Data and sources

• GitHub: the takedown README and star/fork counts for egoist/mygo-cloudmusic (11-06 page: 826 / 369); daily star records from 10-27 onward; the READMEs of egoist/mygo (5,000 stars, 11-06 page) and egoist/mygo-music.
• X: @localhost_4173 10-26 launch post (88.8k views), 11-02 feature post, "got reported" post (55.4k views, 11-04); @QingQ77 (Geek Lite) referral; user replies under the launch post.
• Community: the studygolang introduction post (10-26), and MyGo-related threads on linux.do and V2EX.
• Competitor star counts from GitHub, approximate as of 11-06.
• The actual content of the NetEase letter was not read; the client's real download and user counts are not public.