rw-book-cover

Metadata

Highlights

one of the biggest problems with traditional data analysis. The insights hidden in it like the Isla de Muerta: They can’t be found except by those who already know where they are (View Highlight)

Motif is building a model that takes streams of product event data—which includes the event itself and attributes about the event, like who the user was, the device they were using, how long they’ve been a user, and so on—breaks those events into chunks, and turns those chucks into complex mathematical objects. They then use the same transformer architecture that powers LLMs to predict what events are likely to come next, based on what happened, who did it, and the context in which they did it. It’s a machine for observing everything, and finding patterns (View Highlight)