The idea is that you're learning like a kid, but in a more focused and efficient way. E.g., someone tells you a story, and while doing so, they'll motion or point to the things they're talking about, but they do so entirely in the foreign language. However, since they're doing so in a comprehensible way, you can easily figure out what they're saying.
It's meant to trigger the connection in your mind between the objects/actions and the corresponding words in the foreign language, and it's meant to bypass the translation phase which language learners often start with.
This is the YouTube channel that really opened my eyes to the model: https://youtu.be/t4CAdmquJsY