# A post with 920k views where the top-liked comment is roasting it: a Swiss 21-year-old office worker's motion component library, 47 days to 1,200 stars · GitHub open source project

> Source: https://stringers.io/en/github/gh_kuratlielia__arc-library.html (Stringers / 什么值得做, generated 2026-11-20)
> Stringers tracks which small products suddenly took off each day, investigates why one by one, and distils playbooks from them. Numbers come from 15 databases we collect daily, each with source and date.
> For agents: these investigations and the underlying databases (verified revenue, App Store / Google Play charts, ad libraries, GitHub, Product Hunt…) are queryable live over MCP — tools feed (daily breakouts) / plays (playbooks) / hot / products / data. Setup: https://github.com/komako-workshop/stringers (the user signs in with Google at stringers.io/en/start.html to get a personal URL). Site index: https://stringers.io/llms.txt

> It hit the pain point of 2026 devs fed up with "AI slop" front-end code. The author posts a daily component screen recording on X, 10-09 one controversial post pulled 920k views, and UI-resource roundup accounts and Chinese creators reposted it, so the star count jumped with each viral post.

[GitHub open-source project](https://stringers.io/en/github/index.md) · written 11-9

kuratlielia/arc-library · kuratlielia

# A post with 920k views where the top-liked comment is roasting it: a Swiss 21-year-old office worker's motion component library, 47 days to 1,200 stars

**What it is** A set of React components and page sections with motion, for product UI like dashboards, kanban boards and agent interfaces. You can copy the code, or have Claude Code or Cursor install it into your project with one prompt. Half is free on GitHub, half is paid.

**Why it blew up** It hit on the annoyance of 2026 years, when people using AI to write front-end code kept getting the "AI look". The author posted component screen recordings on X every day, and one controversial post hit 10-09 920k views. Add a UI-resource roundup account and Chinese bloggers reposting, and the stars climbed level by level with each viral post.

**C** **How you could follow this** Stop building React motion libraries. Bring "product UI components for AI that don't look AI-made" to Vue, Svelte, SwiftUI, Flutter, Avalonia and other ecosystems that are still empty. Ship the CLI and MCP on day one, and for every component you build, post a short screen recording on X and ask for stars.

Grade C: the distribution playbook is worth learning, but the React slot got more than a hundred similar libraries in a single month, components can be recreated by AI, and Arc's own post with 920k views only got 1,200 stars. Copying it isn't worth it. Other ecosystems still have room, but you need your own X reach. Public / widely discussed

Data (updated daily) · Repo created 2026-09-24 · 1,328 stars · 7 days +72

GitHub · stars

Data 2026-10-09 → 2026-11-20 (43 days, 42 observations). Points before 10-28 come from deeply / archive backfill, not every day has one

2026 Oct 9: Swiss developer Elia Kuratli posted one line on X: "Claude Code can finally build UI that doesn't look like AI built it," with a link to uiarc.dev and a screen recording of the interface. The post ended up with 920k views, 3,954 likes and 8,189 bookmarks. The most-liked reply (742 likes) was a single line: "your UI looks literally like AI built." Scrolling down, someone said "Performance is poor," someone said "text overlapping text, text over charts," and someone posted a screenshot of broken layout with "great example, idiot." The author replied that it was probably Firefox and he'd fix it. By that night he had added a reply: "The open source version is here (leave a star 👀), and there's a Pro version too." That reply alone got 100k views. Over the next two days, his GitHub repo arc-library gained 264 stars, its steepest stretch so far. Below: what the project is, where the stars came from, and what anyone trying to get stars for an open source component library should copy, and what to avoid.

### What it is

In one line: **a set of React components and page sections with motion, which you can copy, or have Claude Code or Cursor install into your project with one prompt.**

- It follows the shadcn model: `npx shadcn add @uiarc/xxx` installs a component, and an MCP interface lets AI coding tools call the components directly.

- It's not landing-page effects. It targets the UI products use every day: dashboards, charts, kanban boards, signature pads, agent timelines, agent status, tool calls, approval flows. It uses CSS Modules, doesn't depend on Tailwind, and every animation respects reduced-motion settings and supports keyboard use.

- Only half is on GitHub for free: 121 components + 22 sections under MIT. The other half (122 components + 119 sections) is the paid Pro tier, and the README shows Pro placeholders in half of its showcase slots.

- Created 9-24, released 9-29. The author is 21 years old, works as a Solution Engineer at a company during the day and reads at night. It's the only public repo on his GitHub.

### Where the stars came from: every step lines up with an X post

The author built in public on X before launch and has 5,877 followers now (how many he had before launch I couldn't check). His GitHub star count tracks almost one-to-one with his X posts (star counts come from the author's posts and GitHub daily snapshots, view counts are what X displays):

| Timeline | What happened | Stars |
|---|---|---|
| 9-26 | Opened a brand account @uiarcdev, teasing "will ship with MCP" | — |
| 9-29 | Official launch post, 49k views; early-bird codes posted at the same time | — |
| 9-30 | Announced "365 days, 2 components a day." UI resource account @kasuncfdo posted "5 libraries that make your UI look expensive," putting it first, 60k views | — |
| 10-01 | Posted on Product Hunt, didn't make the homepage feed. Someone had posted it to Hacker News the day before, and it didn't take off either | About 170 |
| 10-04 | "Layout morph" component video 44k views; @HoudayfaDesign posted "a UI library you've probably never seen" with 37k views | 10-03 vs 200 |
| 10-09 | The controversial post, 920k views | 380 |
| 10-10 | Signature pad video 53k views; Chinese developer account Yuji reshared it | 483(+103) |
| 10-11 | Kanban board video 59k views | 644 (+161, single-day high) |
| 10-21 to 10-26 | Agent timeline 55k, agent status 110k, tool calls 44k views | 10-27 breaks 1,000 |
| Early November | Still posting videos every day, but no new viral posts | +3 to +10 per day |
| 11-06 | Resource roundup accounts @neropursue (55k views) and @csaba_kissi (17k views) | 1,122 |
| 11-09 | Chinese blogger luinlee posted it, 36.5k views | 11-10 1,205(+68) |

A few things to note:

**The main engine is the author's own daily posts.** "2 new components every day" turns product updates into content: each post is a 10 to 20 second screen recording, titled with the same line "X should feel this smooth", with a link attached. Most videos get a few thousand views, but every few days one hits 50K or 100K, and the ones that blow up are all product UI like agent status, kanban boards and charts. None of them are background effects.

**Hate is traffic too.** Under the 10-09 post, people mocking it as "obviously AI-built" outnumbered the praise, but every rebuttal pushed the post higher. The author didn't delete anything or argue. He just dropped the open-source repo link in the thread and asked people to leave a star.

**Other people's bookmark posts are the second leg.** On X there are accounts that specialize in posts like "X libraries that make your UI look expensive," and Arc was repeatedly added to those roundups from the second day on. After Chinese-speaking programmers like Yuyu and luinlee reposted it, the star count reacted the same day. From the outside you can't tell whether these accounts were paid.

**Launch platforms barely helped.** Product Hunt and Hacker News never took off. The long tail came from directories: the official shadcn registry directory, the MCP plugin page on cursor.directory, and 39 components auto-indexed by 21st.dev (568k views). In mid-October the site also batch-published comparison pages like "Arc vs Magic UI" and "Arc vs Aceternity", and every component page is titled "Animated X for React and shadcn (free)" to grab search terms.

### 920k views earned only 1,200 stars

This is the account anyone trying to copy this should look at closely. One post with 920k views plus a dozen or so videos at 50K to 100K views produced 47 days of 1,205 stars and 53 forks, and it never got onto GitHub Trending. Stars peaked in sync with the X viral posts, then fell back to single digits a day once the posts passed.

The reason isn't hard to guess: component library users want the code, not the repo. Installing goes through `npx shadcn add` or MCP, and copying code can be done entirely on the website, so many people never open GitHub at all. In that luinlee post 11-09, he put it bluntly: dropping a component link into an AI and having it rebuild a copy from scratch in your project's style is often easier than installing a dependency.

For Elia that isn't a problem, because stars were never what he was after. He has said publicly that September revenue was $10.5k, that $3k came in 45 weeks after launch, and that 300 people bought Pro (self-reported, not verified by a third party). GitHub is just the top of the funnel. But if your goal with an open-source library is stars, this playbook's conversion doesn't look good.

### Who's competing: over a hundred new entrants in a single month

| Project | star | How it differs from Arc |
|---|---|---|
| shadcn/ui | 128k | Foundation, components are static; Arc is a registry built on top of it |
| React Bits | 52k | Free flashy visual animations, effects-focused, no full set of product blocks |
| Magic UI | 23k | Landing-page effects (marquees, globes, backgrounds), depends on Tailwind, also has MCP |
| Cult UI / motion-primitives / Animate UI | 4.9K to 7.1K | Animated widgets, few components |
| 21st.dev | 5.6K | Multi-author component marketplace, which ends up distributing Arc for it |

The bigger problem is the copycats that flooded in at the same time. The official shadcn directory now lists 548 registries, 148 of them mention motion or animated in their descriptions, and 117 were added in September and October alone. Newcomers like ns-ui, eased.dev, geturui and cuelark pitch nearly the same selling points as Arc: motion, full block sets, MCP, llms.txt. Someone even 10-10 forked Arc outright, renamed it silicon-ui, listed it in the shadcn directory, and copied the description word for word.

Arc doesn't have any feature rivals can't match. MCP, llms.txt and the registry are all off-the-shelf. It's ahead because of three things: its choice of topics (building the product and agent interfaces AI does worst), its polish, and the author's daily presence on X.

### Switching ecosystems, someone is already doing it

Among the followers who have gotten results, the path so far has been switching ecosystems. The Chinese .NET team AIDotNet built an "Arc-style" Avalonia desktop component library, TKCode.Arc, whose README literally says "Arc-style (uiarc.dev)", and it has picked up 216 stars. The React lane is already crowded, but in Vue, Svelte, SwiftUI, Flutter and Avalonia, complete component libraries that are built for AI use, free of the "AI look", and focused on product UI are still rare.

### If it were you

**Verdict** : there's something to learn here, but copying a React animation component library to chase stars isn't worth it. Arc had a 920k-view viral post and daily videos, and still only got 47 stars in 1,200 days. Meanwhile over a hundred new entrants showed up in the same lane in a month, and the components can be rebuilt by AI from a link. What's worth learning is its distribution method, not its category.

**Who it suits** : front-end developers who already have a few thousand followers on X and are willing to post short videos every day; people who know a UI ecosystem other than React (Vue, Svelte, SwiftUI, Flutter, Avalonia) and know what's missing there.

**Who it doesn't suit** : people who don't want to show up on social media and just want to push code to GitHub and wait for stars. Component library users don't need to open your repo, and without ongoing content, stars stall at a few dozen.

**How to do it:**

- Switch ecosystems: take "product UI components built for AI" beyond React. Prioritize the interfaces AI builds worst, like charts, dashboards, agent timelines and tool calls. The Avalonia version has already shown this path can reach hundreds of stars.

- Take the "AI look" problem and turn it into your selling point: ban purple gradients, glows and cards nested in cards, write those rules into a Skill or prompt file for AI, and ship it in the repo.

- On day one, wire the install path into AI tools (that ecosystem's CLI, MCP, llms.txt), and list it in directories like cursor.directory.

- After each component, record a 10 to 20 second video, use the same title format, and attach the repo link to every one.

- Ask people for stars directly in the README and under the viral posts. Arc's "leave a star" reply got 100k views and was followed by its steepest two days.

**How to get users** : your own daily X posts are the main engine; come up with a slogan that's bold enough to start arguments (something like "finally UI that doesn't look AI-built"), since the backlash will push the post; proactively send your work to X accounts that run "UI resource" roundups and to Chinese tech bloggers; submit to the shadcn directory, 21st.dev and cursor.directory for long-tail traffic. Product Hunt and Hacker News are of little use for this kind of project.

**Timing** : "the AI look in vibe coding" was a hot topic in the second half of 2026, and the React lane saw over a hundred similar projects pile in within a month. The other ecosystems are still empty. If you want in, ship a first version within two or three months, before Arc-type projects launch their own Vue or Svelte versions.

**How you'd know you misjudged it** : if you post daily in a new ecosystem for a month, have a few videos over 10K views, and stars are still stuck at one or two hundred, then this kind of component library is hard to turn into stars and you should change topics. If Arc or a large shadcn-based project ships a version for another ecosystem within a month or two, that gap is closed. Conversely, if a purely open-source animation library like React Bits can reach 10K+ stars in another ecosystem, that path is worth betting on seriously.

[GitHub repo ↗](https://github.com/kuratlielia/arc-library)

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

Competitors

| Name | Where | Launched | Scale | How it differs from this one |
|---|---|---|---|---|
| shadcn/ui (including the official registry/MCP) | Open source | 2023-01 | 127.8K star | It's the foundation, with static components; Arc is a registry built on top of it, with animations and full block sets built in |
| Magic UI | Open source | 2023-06 | 23.2K stars, with Pro and MCP | Does landing-page effects and depends on Tailwind; Arc does product interfaces, uses CSS Modules, and doesn't depend on Tailwind |
| React Bits | Open source | 2024-08 | 51.6K star | Free flashy visual animations, no full product block set, and no paid Pro tier |
| Aceternity UI | Web | Before/after 2023 (unverified) | Not verified | Landing-page showpieces plus paid templates; Arc has comparison pages that target its search terms, focusing on product and agent interfaces |
| 21st.dev | Open source | 2024-09 | 5.6K stars; component marketplace + Magic MCP | It's a multi-author component marketplace; it has already auto-indexed 39 of the author's components (568k views), which ends up distributing Arc for it |
| motion-primitives / Cult UI / Animate UI | Open source | 2024-07 / 2024-05 / 2025-03 | 6.9K / 7.1K / 4.9K star | Animated widgets with few components, no complete product blocks, and no Pro business |
| Concurrent shadcn animation registries (ns-ui, eased.dev, geturui, minidev, cuelark, etc.) | Open source | 2026-09~11 | Of the 548 registries in the official shadcn directory, 148 mention motion/animated, with 9 and October adding 117 new ones respectively | Nearly identical features (animation + MCP/llms.txt); Arc wins on the X buzz it built first, its polish, and the paying customers it already has |

Where users come from

- **Verified** **Author's daily X component demos (365 days, 2 per day)** : @eliakuratli has 5,877 followers; launch post 49k views; Agent status 110k, kanban 59k, signature pad 53k, Agent timeline 55k views; stars +103 and +161 for 10-10/10-11 respectively, in sync with post timing

- **Verified** **Controversial viral post** : 10-09 "Claude Code can finally build UI that doesn't look like AI built it" 922k views, 3,954 likes; top comment "your UI looks literally like AI built" 742 likes; the author's "leave a star" reply in the thread 100k views

- **Verified** **UI resource bookmark accounts / roundup posts on X** : @kasuncfdo 9-30 60k views, @HoudayfaDesign 10-04 37k, @neropursue 10-20 and 11-06 (55k), @csaba_kissi 11-06 17k; reposted on Instagram; whether they were paid is unverified

- **Verified** **Chinese-speaking KOLs** : programmer Yuyu 10-10 (8.5K views), luinlee 11-09 (36.5k views, stars +68 the next day), X 科技时光轴 10-13 tutorial

- **Verified** **Public revenue / build in public** : "I made $10,547 in September" 14k views; "$45,288 in 3 weeks at 21" 13.5k views; "250+ customers 0 refunds"

- **Verified** **Product Hunt / HN** : 10-01 launched on PH, but it isn't in our producthunt database (it didn't get featured); 9-30 was submitted to HN by someone else, and our hackernews database doesn't find it

- **Verified** **Directory sites and programmatic SEO** : the official shadcn registry directory (9-30 listed), the cursor.directory MCP plugin page, 21st.dev auto-indexing, uithings/shoogle/Design Index/Product Rally and others; the site has vs Magic UI/Aceternity/21st.dev comparison pages and a "for React and shadcn (free)" page for each component

- Inferred **MCP / shadcn CLI lets AI agents install it** : claude mcp add ... uiarc.dev/api/mcp; connecting requires registering a free account (2,000+ registered); Pro is also delivered via MCP; actual install counts aren't visible

## kuratlielia/arc-library (Arc / UI Arc, uiarc.dev): why it took off

> Research date 2026-11-10. Repo [github.com/kuratlielia/arc-library](https://github.com/kuratlielia/arc-library), site [uiarc.dev](https://uiarc.dev), X @eliakuratli / @uiarcdev. MIT (free part) + Arc Pro paid.

### Bottom line

Arc isn't an open-source project that went viral on GitHub. It's a **free funnel for a paid component library** . The author, Elia Kuratli, is a 21-year-old Solution Engineer in Switzerland who builds this on the side, working by day and reading by night. He already had a few thousand followers on X and had been building in public. At launch 9-29 shipped the whole package at once: a free open-source core, a Pro one-time purchase, a shadcn registry, and MCP so Claude Code and Cursor can install components directly. After that he posted a daily short video on X showing components, "2 components a day for 365 days." Then 10-09, a post titled "Claude Code can finally build UI that doesn't look like AI built it" (920k views, 3,954 likes, comments in a fight) pushed it to the peak. On X, UI resource-roundup accounts (Nero, KASUN, Houdayfa) and big Chinese accounts (程序员鱼皮, luinlee) amplified it. The GitHub stars came as overflow from that traffic. The differentiator is polished motion in product UI (dashboards, charts, agent timelines, tool calls, kanban boards) rather than landing-page effects, plus the positioning of "built for AI agents, with the AI look removed." That positioning hits the pain point of 2026 year of vibe coding, and it's the main reason this took off. The real moat is thin: the same kind of shadcn registries have sprouted by the hundreds within a month, and someone has already forked it and listed it under a new name.

### 1. Data

- **GitHub stars** (our github library `github history`, from 10-09; before 10-09 we used the author's X posts): 9-24 repo created, 9-29 released; 10-01 about 170 (author's X post), 10-03 passed 200,10-09 380, **10-10 483 (+103), 10-11 644 (+161, single-day high)** , 10-12 about 700. Daily gains then dropped from 20–40 to 3–10 by early November. 10-27 broke 1,000,11-09 1,137, **11-10 1,205 (+68)** . 11-10 GitHub API shows 1,214 stars and 53 forks. Fork/star ratio is 4.4%, and it has never hit trending (the repo is under 60 days old).

- The stargazers list API returns empty, so we can't see who starred it. Fork timestamps cluster between 9-29 and 10-13 (fork list, gh api).

- **Revenue (publicly disclosed by the author)** :
    - September revenue $10,547, of which $6,437 came in a single day, around launch during the early-bird window. Price went up to $199 before 9-30 (X, 2026-10-214k views).
    - 10-19: "3 made $45,288 in a week" (13.5k views), and the same day said "250+ customers, 0 refunds."
    - The official site now says "300+ developers bought Arc Pro". 11-09 The author says the site has 2,000+ registered users.

- **Prices kept climbing** : 9-30 early-bird ended → lifetime $199 / annual $129 (the prices when Product Rally 10-05 listed it) → 10-21 announced "price goes up after 10 more sales" → now lifetime $249, annual $149, team $799 (10 seats; when 10-02 first launched it was $599 one-time / $349 per year).

- **The component count kept growing** : when Product Rally 10-05 listed it, it had 148 components + 86 blocks. 程序员鱼皮 wrote 10-10 with 154+90, a Chinese site wrote 10-13 with 156+90, a Facebook repost wrote 180+107, and now it's 243 components + 141 blocks (free 121+22, Pro 122+119).

- Website: not in Tranco's top 1M (our traffic library has no rank). The author once posted that 10-04 days after launch the site still had 5 people online at the same time, 16.

- Our Hacker News database has no record of uiarc.dev (the author says 9-30 was posted to HN by someone else first, and clearly it didn't take off). The Product Hunt database doesn't have it (10-01 launched on PH but never made featured or the top spots). No record in the indie database.

### 2. Competitors (doing the same thing)

| Name | Where | Launched | Scale | How it differs from Arc |
|---|---|---|---|---|
| shadcn/ui (including the official registry and MCP) | Open source | 2023-01 | 127.8K star | The foundation. Arc is one of the shops on the shadcn registry; shadcn itself is static components with no motion. |
| Magic UI(magicuidesign/magicui) | Open source + Pro | 2023-06 | 23.2K star | Landing-page effects (marquee, globe, backgrounds), built on Tailwind. It also has MCP and llms.txt. Arc focuses on product UI and uses CSS Modules, not Tailwind (per Arc's own comparison page). |
| React Bits(DavidHDev/react-bits) | Open source | 2024-08 | 51.6K star | Free flashy animated components, leaning toward visual effects, no Pro tier. |
| Aceternity UI | Website + Pro | Around 2023 | Scale unverified (closed source) | Landing-page flash + paid templates. Arc has a dedicated comparison page targeting its search terms. |
| 21st.dev(serafimcloud/21st) | Open-source / community marketplace | 2024-09 | 5.6K star | Component marketplace + Magic MCP. It has automatically indexed 39 of the author's components (568k views, 1.5K saves), effectively distributing for Arc. |
| motion-primitives / Cult UI / Animate UI | Open source | 2024-07 / 2024-05 / 2025-03 | 6.9K / 7.1K / 4.9K star | Motion widgets with few components, no complete product-block lineup, and no Pro business. |
| shadcn motion registries that appeared around the same time: ns-ui (389 components + CLI + MCP), eased.dev (motion-first + MCP, 10-19), geturui (free MIT + MCP, 10-22), minidev (500+ free, 10-05), cuelark (agent UI kit, 10-31). | shadcn registry directory | 2026-09~11 | The official shadcn directory lists 548 registries total, and 148 descriptions mention motion/animated; 117 were added in September and October each. | The pitch is nearly identical (motion + blocks + MCP/llms.txt). Arc's edge is the X reach it built first and the paying customer base it already has. |

Verdict: there are a lot of competitors, and a big wave arrived in the month after Arc launched. Functionally it's not unique (MCP, llms.txt, shadcn registry; Magic UI and shadcn had these long before). It took off because of topic choice (product UI / agent UI), polish, and distribution.

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

1. **The author's own X account + daily demo videos (main channel, confirmed)** .
    - Before launch, 9-26 opened a brand account, @uiarcdev, teasing that it would "ship with MCP." 9-29 handed out early-bird codes at early.uiarc.dev. 9-29 the official launch post got 49k views and 508 likes.
    - 9-30 announced "365 days, 2 components a day," and from then on posted nearly every day 2–4 component videos of 10–20 seconds, titled "X should feel this smooth," each linking to uiarc.dev.
    - The hits: 10-04 Layout morph 44k; 10-10 signature pad 53k; 10-11 kanban 59k (these two days lined up with the single-day star jumps of +103 and +161); 10-21 agent timeline 55k; 10-23 agent status 110k; 10-26 tool call 44k; 11-08 pivot table 27k (x library `x user eliakuratli`).
    - His X profile now shows 5,877 followers and 4,597 posts. He says "this is the most successful month since I started on X," which suggests he was already building in public before launch, though his follower count back then is unverified.

2. **A controversial viral post (confirmed)** . 10-09 "Claude Code can finally build UI that doesn't look like AI built it": 922k views, 3,954 likes, 8,189 saves. The most-liked comment was "your UI looks literally like AI built" (742 likes), and a string of comments complained about lag and broken layout on Firefox. The controversy pushed the post higher. In the thread the author added "star the open-source version / Pro for the rest," and that reply alone got 100k views. Stars rose 264 over the next two days, the steepest stretch of the whole run.

3. **Reposts from UI resource curator accounts on X (confirmed; whether they were paid is unknown)** .
    - @kasuncfdo's "5 libraries that make your UI look expensive" ranks uiarc #1, with 60k views and 1,166 likes. 9-30
    - @HoudayfaDesign posted "I found a UI library you probably haven't seen yet" — 10-04 37k views.
    - @neropursue posted the same pitch 10-20 (11k views), then 11-06 put out a roundup post (55k views, 1,248 likes).
    - @csaba_kissi 11-06 compilation, 17k views.
    - Someone also reposted Nero's resource list as an image post on Instagram.
    - Under the author's post, someone commented "check dm or email," which sounds like a promotion pitch. Whether it's paid promotion is unverified. 10-02

4. **Chinese-language circle (confirmed)** .
    - Programmer Yupi posted 10-10 (8.5K views) right at the peak of the star spike.
    - The Chinese site "X科技时光轴" wrote a tutorial (10-13).
    - luinlee posted 11-09 (36.5k views, 497 likes), 11-10 stars +68, timing lines up.
    - The Chinese .NET team AIDotNet built an "Arc-style" Avalonia component library (216 stars).

5. **Product launch platforms (confirmed, small effect)** : 10-01 on Product Hunt. Our library doesn't list it, so it didn't make featured. 9-30 someone else submitted it to HN first, and it didn't take off.

6. **Directory sites / SEO / AI tool directories (confirmed, long tail)** .
    - Listed in the official shadcn registry directory (9-30 first listing, 511 entries sorted by ranking score, ranked #140). Shipped an MCP plugin on cursor.directory, and 21st.dev auto-indexed it.
    - Also listed on uithings, shoogle, Design Index, Product Rally, the webconsulting.at knowledge base, Papyrus, bestaitoolsfinder, WebCatalog, DesignMunk, and others.
    - Starting 10-13, the official site batch-published programmatic SEO pages: comparison pages like "Arc vs Magic UI / Aceternity / 21st.dev" and "Best React component libraries in 2026," with every component page titled "Animated X for React and shadcn (free)."

7. **AI tools install it for users (mechanism confirmed, volume unknown)** .
    - Two install paths: `npx shadcn add @uiarc/xxx`, and `claude mcp add ... https://uiarc.dev/api/mcp`.
    - Connecting MCP requires a free account signup, which quietly turns users into registered users (2,000+). Pro also gets delivered through the same MCP.
    - The author posted a demo of "building a CRM in one sentence with Arc MCP," and that produced a CRM template (10-14).
    - How many installs come through MCP can't be seen from public channels.

### 4. Why it took off

- **Timing** : in 2026, lots of people were writing frontend with Claude Code/Cursor, and the biggest complaint was "AI taste" (purple gradients, glows, cards nested in cards). Arc positioned itself as "the library AI agents use to build interfaces," and shipped a Skill that explicitly bans AI from decorative gradients, glows, and colored shadows (程序员鱼皮's post). That line is easy to explain in one sentence and easy to share.

- **Topic choice** : most competitors do landing-page effects (Magic UI, Aceternity, React Bits). Arc does product and agent interfaces: charts, kanban boards, agent timelines and status, tool calls, approval flows, billing toggles. This is what AI generates worst, developers use it every day, and each piece makes a "wow" moment that cuts well into short video.

- **Distribution runs on the author posting daily** : "365 days, 2 components a day" turns product updates into the content itself, and every post carries a link. The viral 110k and 90k view posts were all component demos, not one-off launches.

- **Public revenue + a 21-year-old day-job dev's story** : "3 weeks, $45,288" and "250 customers, 0 refunds" are exactly the kind of build-in-public posts the crowd loves to repost, and they doubled as social proof for Pro.

- **Open source is just the funnel** : the GitHub repo ships only the free half, and the README has a Pro placement in half its slots. Under viral posts the author himself asks people to "leave a star". Star peaks match X viral posts exactly and then decay fast to single digits a day, which means stars are a byproduct of traffic, not sustained use.

### 5. Differentiation, and whether it caused the spike

- Versus Magic UI / Aceternity / React Bits: **product UI + agent UI** , not landing-page effects; uses CSS Modules + design tokens, **no dependency on Tailwind** ; every animation ships with reduced-motion and keyboard support.

- Versus stock shadcn: animations are built in, plus full sets of blocks (dashboards, settings pages, CRM templates).

- Versus the hundreds of shadcn animation registries from the same period: there's almost no feature gap (MCP and llms.txt are everywhere). The real difference is **polish + the author's daily voice on X + a paying customer base he built first** .

- Bottom line: what actually mattered was "topic (agent / product UI) + polish". Without that, the X demo videos wouldn't have blown up. "Built for AI agents" is more a marketing slogan than a technical moat. luinlee put it bluntly: "Paste the component link into your AI and have it rebuild the thing from scratch in your project's style, often faster than installing the dependency." Component libraries are easy to copy in the AI era.

### 6. Copycats and visibility

- Visibility: high. Revenue was publicly posted by the author on X ($10,547 / $45,288); the launch and viral posts got hundreds of thousands of views, and big accounts in both English and Chinese wrote about it. This is not an obscure project.

- Copycats:
    - teamofsilicons forked the repo 10-10, renamed it silicon-ui, and listed it in the shadcn registry as @silicon-ui. The description is nearly copied word for word ("Free, open source React components and blocks with calm, physical motion... Markdown docs and llms.txt").
    - yuri-xyz forked it as ui-library-arc.
    - AIDotNet/TKCode.Arc calls itself an "Arc-style (uiarc.dev) component library for Avalonia" and has 216 stars.
    - Same-period eased.dev, geturui, cuelark and other "animation + MCP" registries have heavily overlapping selling points.
    - 10-19 someone launched a uiarc token on Base via bankrbot, a name-squatting play.

### 7. Takeaways

- The model is "free open-source core + one-time-purchase Pro + shadcn registry + Pro delivered via MCP". The MCP requires a free account signup, so open-source users quietly become leads.

- Turn product updates into daily short videos, with titles that follow the "X should feel this…" pattern and a link on every post. Whether any one goes viral is luck; the exposure accumulates over time.

- Pick categories AI does worst but that are visual (agent UI, charts, dashboards), and position against "AI-generated look".

- Early-bird codes, limited-time price rises, "price goes up after 10 more sales", and public revenue, all stacked to push purchases.

- Programmatic SEO through comparison pages and component pages, targeting search terms big brands own.

### 8. Still unclear

- The revenue figures are the author's own posts; no third-party verification (none found on TrustMRR).

- We don't know his follower count before launch, so we can't split "existing audience" from "new from this push".

- Whether the resource-collection accounts were paid, unknown.

- Actual MCP and shadcn CLI install counts aren't visible, and how many of the 2,000 registered users actually use it is unknown.

- The stargazer list isn't accessible, so star inflation can't be ruled out. Given how closely stars track X viral posts, it looks more like organic traffic.

- November stars have fallen back to single digits a day; whether Pro templates and daily content can sustain growth remains to be seen.

### Sources

- GitHub API (repo, fork list), our github library `github history kuratlielia/arc-library`

- X: our x library `x user eliakuratli`, `x live "uiarc"`, `x replies 2108492839546147149`; key posts [x.com/eliakuratli/status/2108492839546147149](https://x.com/eliakuratli/status/2108492839546147149), [x.com/eliakuratli/status/2112208153668710902](https://x.com/eliakuratli/status/2112208153668710902), [x.com/eliakuratli/status/2105955466887262446](https://x.com/eliakuratli/status/2105955466887262446), [x.com/kasuncfdo/status/2105153177658306745](https://x.com/kasuncfdo/status/2105153177658306745), [x.com/neropursue/status/2118701420316877011](https://x.com/neropursue/status/2118701420316877011), [x.com/yupi996/status/2108780311761076246](https://x.com/yupi996/status/2108780311761076246), [x.com/luinlee/status/2119613831249100936](https://x.com/luinlee/status/2119613831249100936)

- Official site [uiarc.dev](https://uiarc.dev), /pro, /pricing, /compare/magic-ui

- shadcn registry directory [ui.shadcn.com/r/registries.json](https://ui.shadcn.com/r/registries.json); cursor.directory/plugins/arc-ui; 21st.dev/@kuratlielia; productrally.com/p/arc-ui; time.jamecling.com/bookmark-uiarcdev; motionvault.dev

- Author homepage [eliakuratli.com](https://eliakuratli.com)

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