"IntelliJ, rewritten in Rust": one person built an IDEA alternative in 16 days and got 6,700 stars

l0ng-ai/openidea is a free, open-source desktop IDE that looks and feels like IntelliJ IDEA. The repo was created in Oct 21, reached 6,711 stars by Nov 6, and gained 5,690 in the last 7 days. Nearly all of the code was written with Claude Code. It took off not because of a technical breakthrough but because three things stacked: a number-driven line, "opens projects in 0.9 seconds, 169 MB of memory, where IDEA needs 2 GB"; a Chinese community cold-start playbook the author had run twice; and timing with older users' frustration after JetBrains raised prices and dropped Fleet. Below: what it is, how the stars came in day by day, where it's weak, and whether you should copy it.

Oct 23 morning, an "IDEA alternative" post on V2EX

In 2026, on Oct 23 at 9 a.m. Beijing time, a post appeared on V2EX titled "Wrote an IDEA alternative (open source) so I wouldn't waste tokens":

"I've used IDEA for years and can't give up the feel, but it eats several GB of RAM, the fans roar, and it's just too much. So I decided to build one with the same workflow but lighter."

The post is short: one screenshot of a debugging UI and a GitHub link. By Nov 6, it had 90 replies and the thread was lively. The second reply said "The Fleet JetBrains never shipped could actually work in the AI era." Someone poured cold water: "An IDE's core features are really just two: refactoring and debugging. Everything else has a substitute. Unless you do it better than the alternatives, it's always a toy." Another person tested it: "The IDE takes 600 MB or more, gopls eats 6G, more than GoLand uses."

That same day and over the next two days, the author pushed the project into every Chinese channel he could reach: a long Juejin article, "Rewriting IntelliJ's experience in Rust: project opens in 1 seconds, only 169 MB of memory" (which itself got only 215 reads). He also filed self-recommendation issues for Ruan Yifeng's tech weekly and HelloGitHub, opened a PR to awesome-rust, and posted an English version on dev.to. He'd done this before: his earlier projects tty7 (1,453 stars) and papr (586 stars) also grew by pitching weeklies and HelloGitHub.

What it is

In one line: a lightweight IDE built to mirror IntelliJ IDEA. Double-Shift to search everything, IDEA's macOS and Windows keybindings, and draggable side tool windows that look familiar on open. The feature scope is broad: code completion and navigation (delegated to language servers, claimed to cover 60 languages), breakpoint debugging, Git (graph log, pull requests), a terminal, a database client, an HTTP client, SSH remote development, Vim mode, plugins, and an agent panel that connects directly to Claude Code and Codex.

Under the hood it's a Rust backend wrapped in Tauri 2, with the UI and editor in TypeScript + CodeMirror running in the system WebView. By code volume, TypeScript is about 56% and Rust about 40%. It's Apache-2.0 licensed, has no website, charges nothing, and carries one line at the bottom of the README: "Not affiliated with JetBrains."

It ships absurdly fast: went from v0.1.0 to v0.1.37 in 16 days, nearly two releases a day. The commit history links straight to Claude Code session transcripts, and the V2EX title's "so I wouldn't waste tokens" also says how it was built.

That comparison table

The table sits right at the top of the README, and nearly every repost since has quoted it:

(Author's test: a Python project of about 770 files, on Windows 11 with 32 cores and 62 GB RAM, each run 5 times, median taken.)

Under the table, the author wrote: "IDEA's Python analysis is much deeper than OpenIDEA's, so the 'ready' states aren't doing the same amount of work." He also put the measurement script in the repo so anyone can run it. That line is smart: by admitting the most obvious weak point himself, he leaves critics half as much to say.

How the stars came in, day by day

(Daily GitHub stars, counted by the time each day's scrape ran, which lags the posts by about half a day.)

You can see two waves of amplification: a big Chinese account pushed it from a hundred-odd stars to over a thousand, then a big English account plus Hacker News pushed it to six or seven thousand. After that came a long tail of auto-generated news sites, Medium, YouTube explainers, Reddit, and Telegram channels. I found no sign it ran paid ads or was posted on Product Hunt.

Stars aren't the whole story; downloads line up too. The installers have about 71k cumulative downloads (including auto-updates, so not unique users). The latest release is one day old and has had 7,790 update checks, roughly "a few thousand machines with it installed opened it in one day" (GitHub Releases, 11-06). Windows downloads lead, followed by Apple-silicon Macs. Others have already submitted dozens to a hundred-plus commits, so the community has started contributing.

Why this one

One: the pitch travels, and it has numbers. "IntelliJ, rewritten in Rust: 169 MB vs 2 GB." Reposters don't have to write their own copy; they can paste the whole line. The @GithubProjects post is nearly the first line of the README.

Two: the pain point already exists. IDEA eating memory and indexing slowly is old-user consensus. JetBrains raised subscription prices in July 2025, and in December 2025 announced it was shutting down Fleet, its own lightweight IDE. The "lightweight IDEA" slot was left empty. Under the Nov 6 "uninstalled everything" post, someone wrote "I've subscribed for almost ten years and never thought I'd uninstall the whole suite," and another said "mostly JB's stuff, it eats memory."

Three: it sells familiarity, not novelty. Zed and VS Code can also install IDEA keybindings, but they don't give you IDEA's window layout or the whole set of habits. OpenIDEA lets people skip re-learning muscle memory, something no other option offers.

Four: a wide feature set and fast releases are news in themselves. Debugger, Git graph, database client, and HTTP client all ship on day one, with two releases a day. Reposters call this out specifically. A year ago one person couldn't have done this; now AI-written code makes it possible.

Five: the author has a playbook that works. Repo built, then Chinese developer channels flooded within 48 hours, then wait for a big Chinese account to pick it up, then English aggregators and Hacker News follow. This isn't luck; it's the author's third time doing it.

Its weak spots

The other side of the buzz is skepticism, and it's concentrated:

• "Rewritten in Rust" is a stretch. The top-voted Hacker News comment says the code analysis still runs through an external language server, the UI is TypeScript inside a WebView, and the memory numbers are hard to believe. On X someone replied sarcastically: "it's not rust, it's a webview wrapper tauri." Under @GithubProjects, someone posted a screenshot: "rebuilt with rust >looks inside >tauri."
• Analysis depth falls far short. IDEA's value is in refactoring and code analysis, which language servers can't match. On HN someone said: "IDEA is much smarter than what a language server alone offers." On V2EX someone reported that C# projects won't open their solution, have no highlighting, and can't jump to definitions: "worse than VS Code." On X someone tried it, then uninstalled, saying indexing on large projects was slower than IDEA.
• The name is borderline. IntelliJ IDEA Community Edition was already open source, and critics say "Open IDEA" is misleading. Whether JetBrains will act on the trademark is unknown.
• Whether the positive word-of-mouth is being astroturfed is unclear. Under the "uninstalled everything" thread, someone asked "is this a shill thread? So many people who used it for 2 hours just came to praise openidea?" No conclusion could be reached.

These criticisms haven't stopped the growth; if anything they've kept the discussion hot.

Who else is in the space

(Star counts as of 11-06.)

This table shows two things. The idea isn't rare: at least three other projects pursue the same concept. What differs is distribution. The "lightweight IDEA" slot is already held by a project with 7 thousand stars and another with over 2 thousand, so a third entrant will struggle to get reposted.

How long it lasts

Daily growth has dropped from its peak of 1,629 to 500–600, but it's still rising, and the community has started contributing code. Still, the moat is thin: the technical approach doesn't differ fundamentally from VS Code-style tools, and the hard parts (indexing large projects, refactoring, deep multi-language analysis) haven't been tackled yet. One Hacker News commenter put it bluntly: these "reverse-engineered clone" projects come with "new clones every day." Whether it moves from being reposted to being used every day depends on how it performs on large projects over the next few months.

If it were you

B: a clear playbook worth testing, but don't build another lightweight IDEA. The formula "pick a heavy, expensive old tool that everyone uses daily, keep its habits, rewrite it lightweight, put a comparison table at the top" is copyable, and AI coding lets one person ship a broad feature set. The weaknesses are also clear: the IDEA slot is already taken by OpenIDEA and Lithe-IDEA; "rewritten in Rust" is starting to annoy the English crowd; and a wide feature set means constant bug fixes after launch, or reputation drops faster than stars rise.

Who it suits: developers who have been long-time users of a heavy tool and know which features they use daily versus never; people who can ship thousands of lines a day with AI coding agents and are willing to release daily for months; people with some presence in Chinese developer communities, or who know a couple of open-source recommendation accounts.

Who it doesn't suit: anyone aiming at "lightweight IDEA / PyCharm" (someone already has that, and a third won't get reposted); anyone not planning long-term maintenance (a project this broad dies as soon as updates stop); anyone hoping for technical credibility on Hacker News (these projects get mostly criticized in English-language comment sections; the stars come from reposts).

How to do it, from best to worst:

1. Pick a different heavy, expensive tool: database clients (Navicat, DataGrip), API testing (Postman), Git clients (Sourcetree, GitKraken), C# development (Visual Studio, Rider, exactly where OpenIDEA falls short), Android Studio. Start by searching V2EX and Reddit for "laggy," "eats memory," "subscription too expensive"; the most concentrated complaints are your candidates. Pick the one with the most distinctive interaction habits, so a screenshot alone makes people recognize "this is what X looks like."
2. Build what it does badly: OpenIDEA's weak points are analysis depth and specific languages. Do one language, but do it deeply (say, a lightweight IDE only for Java / Spring, or only for C#), and push it inside that language's community. Lithe-IDEA took this approach, though its stability hasn't kept up.
3. Build one piece: don't build a whole IDE. Replicate only the single most popular window and make it a standalone tool or extension. Someone has already turned IntelliJ's Git client into a VS Code extension (PGshen/vs-idea-gittool, created Oct 17, 236 stars). The engineering scope is much smaller, and maintenance is lighter.

Whichever you choose, copy its packaging: a reproducible comparison table (startup time, memory), with limitations stated under the table by you first; a name that makes clear at a glance who it replaces, plus a line saying "not affiliated with X"; a one-line install, with installers for all three platforms.

How to get users: follow its timeline. Within 48 hours of the repo going up, launch in parallel: V2EX (title written as a personal motivation, like "Couldn't stand X eating memory, so I wrote one"), Juejin long-form posts, self-submissions to Ruan Yifeng's weekly and HelloGitHub, PRs to relevant awesome lists, and an English version on dev.to. Then hand the comparison table and a one-liner directly to Chinese open-source recommendation accounts like jaywcjlove. Once the Chinese circle picks it up, English aggregators (like @GithubProjects) often follow on their own. For the first few weeks, keep releasing daily; the release rhythm itself is material bloggers are willing to repost.

When: as early as possible, ideally before the end of 2026. The "AI solo-clones X" wave is at its peak, and Hacker News is already complaining "new clones every day." Once readers tire of it, the same comparison table won't earn reposts.

Signs you've misjudged it: ① a big Chinese account reposts your project and it's under 200 stars within a week, meaning the formula is already stale; ② OpenIDEA's daily growth falls to double digits by late December and update checks drop to a few hundred, meaning these "clones" can produce a burst of reposts but can't retain users, so reconsider whether it's worth it before you start; ③ JetBrains takes action on the name or trademark against OpenIDEA, meaning riding the name is risky, so keep your project name further from the original product.