I think Musk is on to something. IMO, there are several reasons why it could be a SWB (short wheelbase) two-seater like the ones I have seen depicted on several websites:
1. average cab ride carries 1.4 person
2. cheaper to manufacture, therefore more economical to deploy (also in terms of vehicle weight : kWh needed ratio)
(it will obliterate all those ride-hailers that keep using costly rolling hardware like Waymo's Jaguar iPace)
3. the smaller the vehicle the more road space and agility to evade
4. AV tech experiences less difficulty overviewing the AV's vicinity and the ride process

Having observed my 2023 Model 3 in latest FSD version:
* fail to be able to successfully enter a busy traffic circle
* fail to navigate turns correctly, threatening to hit curbs without intervention
* fail to properly handle unexpected situations such as a tractor trailer poking out into the roadway
I find it highly unlikely that any existing Tesla's will be made to be able to safely autonomously drive without significant retrofit of the hardware. If the camera + compute system can't handle a typical Boston rotary, nobody is gonna ride in these autonomously except in low traffic, simple roadway areas of the US.