I’d maybe reframe in terms of what those points let you do. (The so what?)

Ah, good point. Off the cuff, then: 1. Because each piece of data is in a cell, it's immediately addressable by any other part of the program. Then you don't have to build labyrinths to access parts of your state. 2. You can compose things, like a lego. 3. You don't have to re-run everything all the time. The state is always fresh, always consistent. You get constant feedback for free.

That seems better to me.

1. you can get right into bashing in the most important thing to you: some domain data 2. if you don't want to type dependent data in yourself, just write formulae to do it for you (and to keep doing it when their dependencies change!) 3. you can go on to build on layers of dependencies pointing backwards