Notes on building with strangers

When you build software, you usually don't know your users. This is obvious — but it hits differently when your product is audio and your users are, by definition, ears-first.

They're listening while they drive. While they run. While they do the dishes. They're present and absent at the same time.

The invisibility problem

At a typical software company, you can watch users use your product. With audio, you mostly can't. The moment of use is private, intimate, often solitary.

This changes how you build. You have to build a model in your head — imagining the person driving on the highway at 7am, needing to know if the book is still playing without looking at the phone.

What I've learned

Three things, in no particular order:

The signal is completion, not engagement. A listener who finishes a book has had a fundamentally different experience than one who listened for eight minutes. Completion is slow feedback, but it's honest.

Genre is a proxy for mood. When someone picks a thriller at 10pm on a Tuesday, they're looking for something specific — not literary challenge, not emotional difficulty. They want to be carried.

Silence is data. The books that don't get played. The recommendations that get dismissed. That's the map of where you got it wrong.