BeautifulRust: one "I fell in love with Rust" post, a thousand stars in two days

On the afternoon of 2026 Oct 24, Morgan Linton posted something very personal on X: ten years of C, then a move to Python, always a little homesick for C, until he found Rust and "liked it even more than C." The post ended with a repo: three skills that teach coding agents to write idiomatic Rust. 48 hours later he wrote on LinkedIn, "Over 1,000 stars in 48 hours." By Nov 6 it had 1727 stars. The same thing had been built nine months earlier by someone else, and that top repo still has 1576 stars. How did it end up overtaking it? And could you do it again in another language? This piece lays it out.

What the post said

The middle of the original post reads:

I also think we're moving into a time where we can see code as art in many ways. … In a world where agents write all our code, writing better code is actually even more important imo.

("I think we're entering an era where we can treat code as art … in a world where agents write all our code for us, writing good code matters more.")

No feature list, no screenshots, just an old programmer's feelings, plus a line saying "I distilled Google's Rust style guide, the official API Guidelines and Effective Rust into a skill and am sharing it". The post has so far gotten 74.3k views, 1300 likes, 1100 bookmarks and 153 reposts. Bookmarks are close to likes, which suggests people saved it as a tool rather than just watching.

The author isn't a newcomer: he signed up for X in 2008, has posted 86k times, is co-founder and CTO of Bold Metrics, and is building VulcanBench, an evaluation for coding agents. He already had a readership in the AI-coding crowd (follower count not checked).

What it is

An MIT-licensed Agent Skills package; a single npx skills add morganlinton/BeautifulRust installs it into Claude Code, Codex, Copilot, Cursor, and Gemini CLI. It contains four things:

• rust-best-practices: loads automatically whenever the agent writes Rust, steering it toward idiomatic, maintainable code;
• rust-review: reviews a file, a crate or a diff and produces a report in three tiers (must fix / should fix / consider), with each item marking which rule it breaks and which guideline it comes from;
• rust-port: ports C, C++, Python, Go or JavaScript to Rust, moving the original tests over first to prove behavior matches;
• lints/: a clippy config that puts the machine-checkable rules into CI.

Why this one took off

The idea isn't new. "Installing Rust rules into agents" was already taken at the start of 2026: actionbook/rust-skills (created Jan 17, 1576 stars) is a heavyweight knowledge index, and leonardomso/rust-skills (created Jan 19, 671 stars) is a big table of 26 rules across 265 categories; earlier still there were many small repos with single-digit stars, and in April a Rust review skill called Crabby scored only 6 points on Hacker News.

BeautifulRust won in three places.

The author's own post. The timeline lines up exactly: 10-24 post, 10-25 enters Trendshift's Rust daily chart at #7,10-26, then #4 the same day, +517 stars in one day, past 1,000 in two days. No Hacker News or Reddit posts turned up for it; X is what drove the spike.

The framing hits the Rust crowd's skepticism head-on. Veteran Rust developers distrust "AI rule files." In May, someone on the rust-lang forum asked whether anyone had used a skill to get AI to write idiomatic Rust. Senior member kornel answered: style guides are given, "sometimes completely ignored, sometimes understood too literally, leading the model to rewrite existing code into a warped imitation of the style." BeautifulRust's README answers each point: it lists five sources (plus Microsoft's Pragmatic Rust Guidelines and Rust Design Patterns), says "when sources disagree, pick one and explain why"; includes a before/after of the same agent writing a semver parser with and without the skill (fields go from pub to private, a # Errors doc section is added, examples stop using unwrap()); gives a number: across three test tasks, 30 checks, 29 passed with the skill, 24 without; and promises every example compiles on stable and every lint is checked against the current clippy.

Six releases in a week, each one a fresh post. The next day he posted again: "BeautifulRust is the fastest-growing open source project I've built." The top request in the comments was converting C, C++, and Python to Rust. "I'm building it now." That post got 37.8k views and 926 likes. v1.2 added review (10-25), v1.3 added porting (10-26), v1.4 re-verified every rule (10-26), v1.5 added Bevy and cross-agent evals (10-30).

One more detail: the name. Everything else was called rust-skills; "BeautifulRust" plus a pixel-art crab header is easy to remember and easy to share.

Where users come from

(Daily star gains from Trendshift; totals from GitHub, 11-06.) Directory sites and repost accounts are long-tail followers, not the driver. No mainstream tech press, HN front page, or big X account reposts found.

What happened to the ones that followed

From Oct 28 a batch appeared: rustport-skill (porting-focused), cuda-rust-skill, and several repos literally named rust-skills, all with 0 to 1 stars; in Nov 6 a same-named BeautifulRust showed up too. Nobody has done a "Beautiful [X]" for another language that took off.

Two takeaways: building another one in the Rust niche doesn't get stars; and the content itself (distilling the same public guides) is something anyone can write. What separated it was the author's post and a packaging people trust.

If it were you

B. Don't build another Rust one. The reusable part is the template: pick a language or framework that agents write badly and whose community cares a lot about style. Make a few focused skills + a source list + before/after comparisons + a small eval + a linter config that runs in CI. The template is clear, but getting attention means you have to post and get people to look. This bet depends on your audience.

Who it's for: developers with an audience on X who write code, have real attachment to a language, and have a story to tell; people who know folks in that language's community and know what they hate about AI-written code.

Who it's not for: anyone without a distribution channel who expects GitHub trending or directory sites to carry it (that's how Crabby and the rest of the rust-skills pile stalled at single digits); anyone who just wants to copy the guides into a giant table of hundreds of rules, since the leaders already show that loses to better packaging.

How to do it: candidates include Swift / SwiftUI, Kotlin, Go, modern C++, or a big framework (Bevy is what he added in v1.5). Use it on your own project first, then run the same task with and without the skill and put the difference on the first screen of the README. Name your sources and take a position where they disagree. Every example has to compile, and the lint config has to actually run. Write out install commands for every agent up front.

How to get users: lead the first post with your story with the language, not the features. Over the next few days, pick one request a day from the comments, ship it the same day, release it, and post again. If you hit 1,000 stars within 48 hours, share it once on LinkedIn. This crowd is on X, not HN.

Deadline: once each major agent starts shipping its own official language style packs, this niche narrows. Within a single language, whoever ships a properly packaged version first holds the spot.

How this could be wrong: ① You switch to another language, follow this template, and your first post gets tens of thousands of views, yet you still don't reach 30 stars in 500 days. That would mean BeautifulRust rode Rust's special attachment to "code beauty," not a repeatable playbook. ② Anthropic, OpenAI, or a language's official team starts shipping its own style skill. ③ The other way round: if a Rust follower breaks past 1000 stars by covering what the others missed (say, focused unsafe review or embedded work), then the Rust niche isn't full yet.

Data and sources

• GitHub: morganlinton/BeautifulRust's README and Releases (v1.0 through v1.5); star and fork counts (11-06), tracked daily from 10-25.
• Trendshift (trendshift.io/repositories/320139): +1299 over 14 days; Rust daily chart #7 on 10-25, #4 on 10-26.
• X: @morganlinton's 10-24 first post and 10-25 second post, original text and engagement numbers; 10-26 LinkedIn post.
• rust-lang forum "SKILLS.md for Rust development" (2026-05).
• Competitor star counts from GitHub (actionbook/rust-skills, leonardomso/rust-skills, microsoft/rust-guidelines); Crabby score from Hacker News.
• On 10-26 the daily star reading looks anomalous; the peak date differs between 10-26 and 10-27 depending on the source, so this piece follows Trendshift.