Human-writing repo hits 1,812 stars in two days

A GitHub repo called human-writing picked up 1,812 stars in two days. It's built to make model output sound like one person, per the project's description.

The pace works out to roughly 906 stars a day, real velocity for a writing-and-revision skill with no product behind it. The tool is written for Chinese-language prose, tuned to a specific voice rather than generic fluency, per its description.

Alongside it, a second repo landed: 0xwilliamortiz/humanizer-cli, at 583 stars. It ships as a zero-dependency terminal tool listing 33 signs that text came from a model. Run it as a draft-checker before you publish anything a model touched, per its repo description.

Together, the two are trending as anti-slop tooling turns into its own category on GitHub. Neither changes what the writing says. Both work on how it sounds: rhythm, word choice, the texture that makes a sentence read as one person, not an average of training data.

One repo teaches a model to write less like a model. The other teaches a human to catch when it didn't.