One tweet, 440 stars in two days: Anthony Fu wrote a sequel to i18n-ally, a tool he dropped five years ago, x18n

In 2026 Oct 28, Anthony Fu, one of the best-known open-source authors in the Vue / Vite community, posted a tweet introducing his new i18n tool, x18n. Within two days the repo went from 29 stars to 473. Why it took off came half from the author himself and half from the VS Code extension he built 2019 years ago, i18n-ally, which has 1.05M installs and has been unmaintained for nearly two years. Below: where the stars came from, which parts you can learn from, which you can't, and whether it's worth trying to get stars in this space.

One tweet, plus a line calling it a "sequel"

10-28 at noon 12:37 (Beijing time), @antfu7 posted:

🌏 Introduce x18n, the ultimate toolkit for managing i18n files! You get CLI, linter, interactive UI, LSP, editor support, CI/CD, and native agent integrations… For any frameworks!

A few minutes later he quoted his own post and added: 「If you ever enjoyed i18n-ally, the VS Code extension, this is the spiritual successor to it!」 (If you ever liked i18n-ally, the VS Code extension, this is its spiritual successor!), plus a note that the migration guide and scripts were ready.

The first tweet reached 63.5k views, 1,430 likes and 140 reposts at 11-06; the follow-up had 150 likes. The repo was created 10-15 and had been tested on a small scale for two weeks, with only about twenty stars before the post.

(GitHub daily stars, 11-06)

The jump in stars over those two days lines up exactly with the tweet. Later reposts were small: a Chinese aggregator account wrote "Anthony Fu open-sourced another new project" (3 likes), and a Japanese account and an Angular developer said they wanted angular/localize support (4 likes each). I found no traction on HN, Reddit or Product Hunt. It's fair to say the stars all came from that tweet.

What it is, and its relationship to i18n-ally

Anyone who has done multilingual frontend work knows the pain: translations scattered across JSON and YAML files, missing keys, figuring out which text maps to which key, and how many files you have to touch to rename a key, all done by hand. i18n-ally solved this: inside VS Code it shows translations right next to your code, flags missing ones, and lets you edit them with a click.

Anthony Fu wrote it in 2019 and handed it over to the translation platform Lokalise in 2021. Updates became less frequent, and the last commit was 2024-12. In its issues section, people were asking "Is this repo still alive?" as early as 2022; in February 2026 someone forked it as i18n-ally-next, announcing "A Maintained Fork" and getting 27 reactions. People have been looking for a replacement all along.

x18n was rewritten from scratch. On the official "Why x18n" page he says he had long wanted to split i18n-ally's core out of VS Code, but the extension was already built into VS Code's interfaces and couldn't be pulled apart. Then:

Agents make that possible. A rewrite of this size used to need months, and now I can build it in my free time.

(Only possible with agents. A rewrite this big used to take months; now it can be done in spare time.) The commit history does show many PRs signed "via agent" and by the Cursor agent: 22 days, 247 commits, and 29 tags.

After the rewrite, one core powers: the command line, lint in CI, a browser management UI, a language server (any editor can connect), VS Code and Zed extensions, and Vite/Nuxt DevTools panels. It ships with presets for 45+ frameworks (Vue, React, Angular, Svelte, Next, Nuxt, Flutter, Laravel, Rails, .NET) and reads JSON, YAML, PO, Fluent, resx and other formats. It also leaves a deliberate opening for agents: stable --json output, reverse lookup of keys from text, and safe key renaming. The official site's line is "they edit locale files all day": agents spend their whole day editing translation files.

Beyond stars: is anyone actually using it?

Two things stand out. First, command-line downloads kept climbing after the tweet buzz faded, doubling in a week, which suggests people have wired it into CI or daily scripts (npm numbers include mirror and repeated CI pulls, so discount them). Second, Open VSX downloads are 17 times the official marketplace's. Open VSX is the extension source used by AI editors like Cursor and Windsurf. Early users were mostly in AI editors, which fits the "built for agents" pitch.

The ecosystem is picking it up too: Nuxt lead Daniel Roe's diglot project shipped the @diglot/x18n plugin 11-02; a community member tweeted that they were added to the x18n-dev org 11-03.

On the other hand, i18next's own command-line tool, i18next-cli, has only 251 stars but 1.14M weekly downloads. In this space, stars and actual usage are largely unrelated; x18n's 684 stars mostly reflect attention earned from reputation.

Who it's competing with

(GitHub, npm, extension marketplaces, 11-06)

i18n tools are not in short supply. What's missing is the one that's framework-agnostic and still has an editor experience. That spot used to belong to i18n-ally, and now its original author has come back to pick it up himself.

What's learnable and what isn't

Not learnable: Anthony Fu has 62k followers on X and 40k on GitHub, and he's a core author of Vite, Vue and Nuxt. i18n-ally was also written by him, so the "successor" framing only lands coming from the original author. Followers can't copy that.

Learnable: the cold-start playbook.

1. Find a widely installed tool whose author has left or that has stopped getting updates, then build its successor. The old user base is already there, and "is this still alive?" posts in the issue tracker are proof of demand.
2. Make migration a single command, and prompt users to run it. x18n has x18n migrate i18n-ally, and the VS Code extension detects i18n-ally settings in a project and asks if you want to migrate.
3. Separate the core from the editor. One engine serves the CLI, CI, every editor and agents. Old tools can't do this, and a new project can make it its pitch.
4. Say "for agents" in the pitch (JSON output, skills, MCP). Publish the extension to Open VSX too, to reach Cursor users.

If it were you

Verdict: C. You can learn from it, but following it isn't worth much. Nearly all of this project's stars came from its author's one tweet and the "original author returns with a successor" story, and you have neither. The i18n slot is already taken, and building another general-purpose i18n tool would struggle to pull attention away from Anthony Fu.

Good fit for: developers who have deep hands-on experience with a stalled older tool and can make its migration a single command; or people with connections in an ecosystem x18n doesn't cover.

Bad fit for: anyone who wants to head-to-head build another general i18n toolkit; anyone without a posting channel who is counting on organic GitHub traffic (a new repo takes about 60 days to reach trending, so stars have to come from posts).

How to do it, ranked by how worthwhile:

1. Copy the playbook in another domain: look in the VS Code Marketplace and npm for tools with hundreds of thousands of downloads, no update in over a year, and issue threads asking "are you still maintained?". Build the "not just for VS Code" successor, with a built-in migration command, old-config detection and --json output. This is the most useful thing to take from this piece.
2. Fill the gaps x18n left: for example, the angular/localize support people asked for in replies, plugin inference for JetBrains-family editors, or deep presets for a specific framework. Send PRs to x18n or build an ecosystem plugin (like @diglot/x18n). That builds reputation faster than starting a standalone repo.
3. Building "another successor to i18n-ally" isn't recommended.

How to get users: write a post with the same structure as that tweet: first line says what it is, second line names the old tool ("if you used XX, this is its successor"), third line says migration is one command. Post it on X, and reply under the "still alive?" threads in the old tool's issue tracker. The people in those threads are your first users.

Timing: there's no time window for the stalled-tool route, but the idea that "agents let one person rewrite a large tool in their spare time" is fresh right now. The pitch works best before the first half of 2027.

Signs you've misread this: ① If x18n's daily new stars drop back to single digits after November, and weekly CLI downloads fall back to a few thousand, then even a famous author plus a successor only gets one wave, and the ceiling for this playbook is lower than it looks. ② The reverse: if an unknown author gets several hundred stars in two weeks with "successor + one-click migration", the playbook works without the fame, and the first route should move faster. ③ If Lokalise starts maintaining i18n-ally again, old users have no reason to move.

Data and sources

GitHub: x18n-dev/x18n repo and daily stars (10-24 to 11-06), commit history; lokalise/i18n-ally issue #868 (2022-12), #1376 (2026-02). X: @antfu7 two launch posts (views and likes as of 2026-10-28 11-06), posts by @meow2149, @yug1224, @erkamyaman_ng, @0froQ. x18n.dev homepage "Why x18n" and migration guide. npm download counts (as of 11-05). VS Code Marketplace and Open VSX installs and downloads (11-06). Where this says "inference", it is inference.