A YouTube channel with 1,320 subscribers, one 89-second video: How Raven got nearly 2,000 stars in a month with a "C-language GUI library"

Raven is a desktop GUI library built by Oscar Soirey, a developer in Montreal, working alone. It's written in C, rendered with Vulkan, looks like macOS Liquid Glass, defines interfaces in XML, and can be called from ten languages. Nearly all its stars came from one demo video on the author's own channel, which YouTube recommended to about 300k views. Nobody on HN, X, or tech media has written about it. The playbook works with any topic, but building another "beautiful C GUI library" on top of it wouldn't add much.

A video the algorithm pushed

On Oct 22, 2026, Oscar Soirey posted a 1-minute-29-second video on his YouTube channel titled "Raven - An Open Source Modern GUI Library Built with C & Vulkan." The channel had existed since 2018, with only 6 videos and 1,320 subscribers.

On screen: frosted-glass panels, rounded buttons that squish when pressed and spring back, and pop-ups that burst out from the point you clicked. The interfaces C programmers usually see are gray boxes like Dear ImGui. The video looked like an iOS app.

Three days later, this comment appeared (366 likes):

"YouTube started recommending you, congrats. I found this through the recommendations."

By Nov 6, the video had 299,729 views, 6,044 likes, and 451 comments (YouTube, 11-06). The channel had only 338k total views, meaning almost all its traffic came from this single video.

Why the stars arrived five days late

When the main video was being pushed, the GitHub repo had no source code. On Oct 25, the author pinned a note under the video (599 likes):

"The project was originally going to be split into a free and a paid version, so the source wasn't in the repo. It's now fully open source, and that's been corrected."

The star trajectory matches this:

Likely the traffic arrived around Oct 25, but with nothing in the repo to look at, stars didn't climb. Once the source was added, the accumulated attention turned into 700 stars within a day. I couldn't verify exactly when the source was added. Either way, several days of traffic were wasted.

Other channels are almost silent: no Hacker News post, no discussion on X, only a Reddit repost of the 0.2.0 video in r/coolgithubprojects. A comment says "waiting for Fireship to make a video," so the big accounts haven't picked it up yet. The next two videos got fewer views each time, and daily stars dropped from about 250 to about 33.

What it hits

1. It looks good. Frosted glass (clear, frosted, or tinted), spring animations, and it can mimic the look of Windows 11, macOS, and GNOME. C and C++ GUI libraries have been engineer-designed for decades, and commenters say things like "finally, we waited decades for this" and "I've been looking for this for ages."
2. "Written in C, zero dependencies." The README says C11, it opens its own window, and it loads Vulkan at runtime. The stripped shared library is about 1.3 MB. That label alone draws the C crowd. A comment with 173 likes puts it next to Clay: "First Clay, now Raven, both zero-dependency. This is a good year for native GUI."
3. One XML file, ten languages. C, C++, Rust, Python, Go, C#, Zig, LuaJIT, Java, and Odin can all load the same UI file, and the interface refreshes on save. It also ships a visual designer and a theme editor. Communities of languages with few GUI options showed up: "Zig doesn't have a good UI framework right now" and "I want to see this in Odin's official libraries."
4. The criticism is spreading it too. The top comment was "I'll wait for Tsoding to try it" (837 likes), followed by "vibe-coded liquid ass, now C has it too" (313 likes) and "this looks so vibe coded" (178 likes). The author, Oct 31, opened issue #15 and publicly said that window management, most of the Vulkan implementation, the XML parser, the binding generation scripts, and the commit messages were written by AI, while he handled the API design and architecture. The stars kept climbing after the confession.

Point 4 also explains why it spread so fast: under a month of development, about 226 commits and 7 releases. The windowing layer covers Win32, X11, Wayland, and Cocoa. Bindings for ten languages were published to PyPI, NuGet, crates.io, and similar package registries, along with node graphs, timelines, docking panels, and a code editor. One person writing this by hand isn't realistic. The feature list is long enough to make people star it before they've even tried it.

Who else is doing the same thing

The space is crowded, and Raven has nothing unique on features: Slint already does declarative plus multi-language plus designer. What's unique is the combination: C, a new Apple-style design, one XML file callable from many languages, and a video you understand in 90 seconds. I haven't seen anyone start a separate "glass-style C GUI" from scratch; everything I can find is a direct fork.

Its weak spots

• Stars aren't usage. I couldn't find package download numbers, and only 6 issues are open. Most of the people starring it probably just watched the video and thought it looked nice.
• People doubt it will last. The top comment with 310 likes was "betting it gets thrown away in two months." The author doesn't fully understand the AI-written code, the API is still changing, and anyone who'd actually build a project on it will hesitate.
• Growth came from one recommendation. The second and third videos got fewer views each time. Once the algorithm stopped pushing, the curve flattened.
• Name clash. Bottles' author Oct 23 also released a project called Raven (running Windows apps on Linux) at nearly the same time, so search results get mixed up.

If it were you

Bottom line: Copy the playbook, not the topic. Pick a developer category whose looks have been ignored for a long time, use AI to make it good-looking and complete, then shoot a 90-second video. That formula works with other topics too. But building another "beautiful C GUI library" will be seen as a knockoff, and you'll have to compete head-on with Slint and ImGui.

Who it suits: Developers with a design sense who are willing to polish the demo, where the quality is obvious at a glance: UI libraries, terminal tools, editor themes, visualization, animation libraries.

Who it doesn't suit: People building pure backend, CLI, or protocol work that doesn't produce anything visual; people who can't handle comments calling their work "vibe coded"; people who want to earn money from open source. Libraries like this rarely make money. What you get is reputation and followers.

How to do it:

1. Find something many developers have used for years but that has never looked good: a language's TUI framework, a game engine's editor UI, Godot or Unity interface themes, a language's charting library. Raven proved in C that "looks good" alone can earn stars.
2. If you want to build it in another language, go where Raven's comments named a missing UI framework: Zig and Odin. You could also fill the gaps it left: a stable API, complete docs, clear labeling of what AI wrote and what was tested. That's for the people who are afraid it will be abandoned in two months.
3. Before posting the video, put the source, runnable examples, and prebuilt packages in the repo. Raven lost several days of traffic that way.
4. State up front how much AI was used. Raven only opened issue #15 after being criticized. Disclosing early is better.

How to get users: Post a 60 to 90-second video on your own YouTube channel. Show the best visuals in the first 5 seconds, and make the title say "language + what it does + what's different," like Raven's "Built with C & Vulkan." A small channel is fine; Raven had only 1,320 subscribers. Within a week, follow up with a tutorial video to catch new viewers, and put the same GIF and video link at the top of the README. HN and Reddit are worth posting to, but don't count on them.

Why it has to happen now: The "native GUI" wave is still hot. Clay, Raven, and those two Rust frameworks all appeared this past year. My guess (inference) is the window is a few months. Move fast, before big channels like Fireship's cover the topic to death.

What would show you misjudged it: If the video sits at a few thousand views two weeks after posting and never gets recommended, the visuals aren't striking enough, or this niche isn't on YouTube. Raven itself stalled around two thousand stars in mid-November, and comments started saying "no longer maintained." That shows the stars from "looks good" fade faster than expected, and that the formula is hard to copy with a different topic.