The app was built by Abelardo Riojas, a data science graduate student and lifelong music listener, who curated a dataset of textual song metadata and trained a model to describe songs by their musical and cultural traits. That dataset is what lets the search engine reason about genre, lyrical meaning, sonics, mood and energy, so a prompt like 'Midwest emo songs to cry in the shower to' returns a coherent set of tracks rather than a random assortment.
Using it is deliberately simple. You describe the playlist you want in a sentence or two, and the model returns matching songs. Logging in with Spotify writes the generated playlist straight onto your account and feeds your listening back into improving the algorithm; a guest option is available if you just want to experiment. Guidance on the site recommends leaning on adjectives and cultural context rather than artist names, and using positive phrasing because the model struggles with negations.
Because it is a personal, independent project rather than a commercial streaming service, the tool is free to use and focused on discovery for dedicated music fans. It is best understood as an alternative way to find music: instead of clicking through genres and tags, you describe the feeling you are after and let language-powered search do the digging.
