I'm guessing their initial target is going to be more like a bus service in terms of routes. Better than a bus in having fewer passengers (and thus fewer stops), and more precise pickup/dropoff locations. But probably not door to door like a taxi service, at least not unless your start and end point are both within their geofence, which I guess will occur in urban areas.

In some limited areas they can probably get something up and running fairly quickly compared to the challenge of building a robotaxi (and it looks like they are, in San Francisco). Can't handle unprotected left turns very well? Route around them. Can't handle many parking lots? Don't allow pickup/dropoff there.

I wonder how they're doing testing. Could have someone with a laptop plugged in and able to "make strategic decisions," as Waymo calls it.

Do they even call it a taxi?