freeCodeCamp at 437.9k stars: the one repo out-starring React that isn't a scam

What happened

The GitHub trending page today has an unusual top three. freeCodeCamp/freeCodeCamp sits at 437.9k stars. Below it, openclaw/openclaw — the lobster-emoji repo with a one-line README that we covered five hours ago — sits at 283.1k. Below that, facebook/react at 243.9k.

That ordering is not a bug. freeCodeCamp has been the highest-starred repo on GitHub for years; it briefly traded the top slot with `sindresorhus/awesome` in 2021 and hasn't given it back. What's new is the neighbors. A twelve-year-old nonprofit curriculum, a five-day-old scam repo, and the third most-depended-on JavaScript library in existence are now sorted next to each other on the same leaderboard, with nothing but a star count between them.

If a hiring manager, an investor, or an LLM ranking training data can't tell those three repos apart from their GitHub badges alone, the badges have stopped doing the job they were designed for.

Why it matters

We've written about star-count decay four times in the last two weeks — the awesome-python piece 18 hours ago made the point that a curated markdown file is out-starring every library it links to. This isn't that story again. This is the inverse: freeCodeCamp is the one repo at the top of the leaderboard that arguably *deserves* to be there, and looking at why is more useful than another round of pointing at broken metrics.

freeCodeCamp is not a library. Nothing imports it. Its `package.json` is an internal build config for the curriculum site, not a distributable. It ships no npm package, no Docker image, no binary. By every technical definition of "software project," it's a website with lesson plans in a Git repo. And yet it has almost twice React's stars.

The reason is that freeCodeCamp is the only repo in the top ten where the star is a real endorsement. Roughly 10 million learners have gone through the curriculum since 2014. The public certifications page lists over 40,000 people who've completed the full-stack track. A significant fraction of those learners star the repo the day they finish, the way a graduate signs a yearbook. The star is a receipt for something the person actually did.

Contrast openclaw. Its 283k stars accumulated over five days from a coordinated pattern we documented four hours ago — accounts starring dozens of freshly-created repos in the same window, README containing nothing but a lobster emoji and a marketing tagline. The stars are real GitHub API events. The endorsement is fake.

Contrast React. Its 243k stars accumulated over twelve years from developers who imported it, shipped it, and clicked star as a bookmark. The endorsement is real but the *unit* is different: one star per developer-decade, not per graduation. React's growth is capped by the population of people who write JavaScript for a living and remember to star things. freeCodeCamp's growth is capped by the population of people learning to code, which is orders of magnitude larger.

Three repos, three completely different underlying phenomena, one identical metric. The community reactions on HN in the last 48 hours have converged on the same conclusion from different directions: `simonw` noted last week that he now sorts trending by "created date > 2 years ago" as a first-pass filter; `dang` has been quietly downranking submissions whose only claim is a star count; the freeCodeCamp maintainers themselves stopped mentioning the number in their annual reports around 2022.

What this means for your stack

If you use GitHub stars as a signal in an automated pipeline — dependency scoring, tech-radar generation, LLM training-set curation, procurement due diligence — you now have to segment the number before it means anything. Curriculum repos, awesome-lists, and prompt-injection bait repos all live above 200k stars and none of them are software. A flat `stars > N` threshold is now producing false positives at the top of the distribution, not the bottom.

The cheap fix is a three-line filter: exclude repos where the primary language is Markdown or HTML, exclude repos with fewer than 100 lines of actual source, and exclude repos created in the last 30 days with star velocity above some threshold. That knocks out freeCodeCamp, awesome-python, and openclaw respectively — leaving you with something closer to the pre-2022 signal.

The better fix is to stop treating stars as a scalar and start treating them as a distribution over time. React's stars are a slow accumulation with a long tail. openclaw's are a five-day vertical line. freeCodeCamp's are a decade-long steady climb tied to curriculum completion cohorts. GitHub exposes the per-day star history via the API; three joined columns and a slope calculation gives you a signal that survives all three failure modes at once. Nobody's shipping it as a default view — but if you're the one writing the LLM-agent tool that reads GitHub for its rankings, you're the one who has to.

Looking ahead

freeCodeCamp being at #1 is the least interesting fact about today's trending page. The interesting fact is that the same leaderboard now surfaces educational infrastructure, adversarial bait, and production libraries with equal visual weight, and GitHub hasn't shipped anything to distinguish them. Expect that gap to be filled either by GitHub itself (a "verified library" badge is the obvious move) or by a third-party ranking layer that becomes the new default for anyone doing automated repo evaluation. Whichever ships first inherits a lot of what stars used to mean.