"Can't git clone": how a single complaint turned into a global law repo with 1,800 stars

A small legal AI company posted a long Hacker News article about "git for the US Code" and it hit #1. The top comment said the title was fraudulent because there was nothing to clone. Within 12 hours, using Claude Code, he built an open source "all the world's laws" repo and came back to the thread with the link. Two days later a big X account reposted it, and the repo gained 880 stars in one day. This piece covers what lexifina/global-law is, where its stars came from, and what someone who wants stars for their own open source project can copy from it, and what they can't.

One bad review

On Oct 28 2026 01:47 (UTC), user alansaber posted "Git for the US Code" on Hacker News, linking to a post on his company Lexifina's blog. The article describes turning the US Code into git history: each law taking effect is one commit, so you can see what any section looked like on any day and trace each word to the law that inserted it. The piece walks through the methodology, noting that 361 of 365 amendments could be reproduced word for word, and uses the COVID-era CARES Act raising bankruptcy limits as an example of why timing matters. The post reached 454 points and 267 comments that day and ranked #1 (HN, 11-06).

The top comment, from user octoberfranklin, was just one line:

Fraudulent title, you can't git-clone this.

Others said they couldn't find anywhere to try it, and some argued that this kind of thing shouldn't be done by a private company and should go into the public domain.

On 13:34 (UTC), lexifina/global-law was created on GitHub. The next morning, alansaber replied under that complaint with "Actually, you can now", linked the repo, and posted a few more comments in the thread. He also submitted a new post, "Git for All Global Law", which got only 1 points. Its only reply was "you already had a successful post yesterday" (HN, 11-06). The new post didn't take off, and the stars came from the comments on the original post and the reposts that followed.

What it is

A repo that turns national laws into git history, licensed CC0.

• Each country gets its own repo, lexifina/<country>-law, with one commit per law taking effect, and the commit author is the law itself. global-law uses submodules to string them together into a single global timeline.
• It also ships law.py, which can show the text of a section on a given day, blame which law wrote a given line, and diff versions. There's also a read-only MCP server with 18 tools that makes no network calls. The repo already includes .claude, .cursor, .vscode, and .codex configs, so after cloning, Claude Code or Cursor can look up statutes directly.
• The repo itself says it was built in one day on Oct 28 with Claude Code: "design, code, and prose all produced by the model, with no changes to the legal text," and "not reviewed line by line by a lawyer" (REPRODUCE.md, quoted by LockerGnome).

It has grown fast. The 11-06 main repo has over 340k commits and 15k tags. The lexifina account has 700 public repos, most of them country, state, and city law repos created in bulk between 10-28 and 11-02 (GitHub, 11-06). Coverage is uneven: when LockerGnome wrote about it (10-31), Czechia, Spain, and Switzerland had entire codes, the US had only 9 titles, and the UK had a single special education law. Checking again 11-06, the China repo already has 172 national laws, and the criminal law has been committed amendment by amendment from the 1997 revision through Amendment (XII). It is still being filled in country by country with the model.

How the stars came

Daily new stars (trendshift, as of 11-06):

Total: 1,834 stars and 173 forks on GitHub (11-06).

HN gave it credibility, not stars. On the day the article hit #1, the repo didn't exist yet. Afterward, nearly every repost mentioned that it had "ranked #1 on Hacker News."

Most stars came from one X post. The original post by @RoundtableSpace on 10-30 09:45 (UTC) read:

Someone is putting the law of every country into one git history… So you can see exactly what the law said on any date, and which law changed every word. It started with the US Code, which hit #1 on Hacker News… Free to use for anything, and perfect for feeding to an AI.

The post has 250k views, 4.3K likes, and 4.2K bookmarks (X, 11-06). Bookmarks and likes are close in number, which suggests many people saved it to feed to an AI later. That single day's +880 accounted for nearly half of the total stars. This account often posts "someone built an X" tech reposts; whether this one was paid promotion can't be verified.

Then the repost chain. LockerGnome's long article and Chris Pirillo's own X post came next, followed by LinkedIn posts saying "don't ever show this repo to a lawyer" and AI news accounts on Threads and Instagram. The headlines were all similar: "Free Westlaw alternative," "All the world's laws, perfect for AI." The author also posted "Using Claude to put global law into Git" on r/ClaudeAI. Its engagement isn't visible.

The fork rate is about 9.4%, which looks high, but the README says "fork it, ship it," and the submodule structure invites casual forking, so it shouldn't be read as proof many people are using it.

Who else is doing the same thing

"Putting laws in git" isn't a new idea.

Two comparisons are worth noting. First, us-code: nearly the same thing, also on HN, but only reached 42 points and 374 stars. The difference was the article and the story. Second, legalize: the Spanish repo alone got close to 2,000 stars, but the same person's German and French repos never took off. So stars don't track how many countries you cover. They track which single moment gets seen by a lot of people.

What global-law adds beyond those is three things: the "whole world" story, word-level tracing back to a specific law, and an MCP that plugs straight into AI. The first one drove the stars; the other two gave people a reason to keep looking after they clicked.

Why it works

1. Publish a solid long post first, then the code. Methodology, reproduction rates, and concrete examples are what get you to #1 on HN, and that endorsement is what every repost leaned on afterward.
2. Fix the top complaint the same day, and make the story bigger. When the comments said "can't clone" and "shouldn't be private," the author released a CC0 open version the next day, and expanded from one country to the whole world.
3. Make it a one-liner people can repost. "All the world's laws in one git repo" and "feed it to AI" needed almost no rewriting for repost accounts.
4. Ride a real AI problem. The long post cited research showing models cite repealed old statutes, and accuracy rose from 24% to 87% with retrieval. "Any date" plus MCP hits that problem directly.
5. Use the model to fill out coverage fast. Creating hundreds of country and region repos in a day is what made "the whole world" credible; then they filled in each country one by one.

If it were you

C. The playbook is worth learning, but building another "legal git repo" isn't worth it. The model is already filling in country by country, and Chinese, Spanish, German, and similar slots are already taken.

One-line take: what you can copy is the shape (vertical data in git + MCP + a long post that can reach #1 on HN), not the legal topic. The shape alone isn't enough: us-code did much the same thing and only got 374 stars.

Who it's for: developers who have a dataset nobody has organized, that changes over time, and that AI often gets wrong, and who can write a long, methodical article about it. Examples: tax rates and tax law by country, industry standards beyond ISO/IETF, drug label revisions over time, public company articles of association, open-source license version histories, and the revision history of platform developer agreements and privacy policies.

Who it's not for: people who want to redo the law library for another country (lexifina fills that faster, and already owns the "all countries" story); people who can set up a repo but won't write long-form posts (no article, no HN #1, no big account willing to repost); and people who want to build a commercial product on it (CC0 data, and the moat isn't in the data).

How to do it: pick one domain and first make one subject reproducible word for word, for example "every revision of a platform's developer agreement from v1 to today." One commit per revision, with blame and a read-only MCP, and put .claude and .cursor configs straight in the repo. Make the README's first screen say clearly what's covered and what isn't, before someone else points it out. Then use models to batch out the remaining subjects.

How to get users: follow its order. First publish the long article, built around one concrete "point in time" example (the day a clause changed, and how AI answered wrong before that). When you post to HN, the repo has to clone; don't wait for the comments to call you out. If the comments raise complaints, fix them the same day and post the link back. Then send a one-line version to X and LinkedIn accounts that specialize in "someone built X" posts. The Chinese version can follow the same order on V2EX and Jike.

When to move: don't rush. It's not a meme. The shape stays valid for a long time. But while "git + MCP, feed it to AI" still sounds new, it's best to ride it, ideally before the first half of 2027 is over.

Signs you've misread it: ① In November, two or three more "some dataset in git + MCP" repos appear, each also getting over a thousand stars. That would show the shape can earn stars on its own, without HN #1 or big accounts, so act right away and raise the grade. ② The reverse: repos like global-law where nobody really uses it (issues stuck in single digits, commits stopped in November). That would mean the stars came from bookmarkers, and following up only gets one-off stars. ③ Westlaw, LexisNexis, or a model vendor ships an official "look up statutes by date" MCP. Then this category has no new story left to tell.