Show HN: I used n-grams to see which LLM writes most like a human

Hi HN,

After learning about n-gram models in a Natural Language Processing class, I got curious about how well they could represent something abstract like writing style. To test this, I built a simple game that compares your writing to that of three LLMs: GPT-4o, Gemini 2.5 Flash, and Claude 3.7 Sonnet.

You write a short story based on a prompt, and the tool uses pre-built n-gram models to tell you which LLM's style is most similar to yours. There's no sign-up required to try it.

I wrote a companion post that goes into more detail on how the comparison works and how I built the models: https://ramjanarthan.com/teaching-computers-language/

I'm keen to hear any feedback you have.