Not wanting to fuel the flames, but that must be applied to several of the solutions. Eg some solutions are speed and/or scenario limited, not because of technical ability, but because that is what they are either licensed for and/or comfortable airing in public. Tesla seem comfortable doing their testing in public, other solutions less so. It doesn’t mean that is not happening, just that progress and full capabilities are less obvious or the capabilities, no matter what limits some may think they have, are to meet more specific objectives and meet those well.
The Mercedes example described earlier is a good example of this. Their system can drive at higher speeds and without a lead car when operating at level 2, however a combination of Germanic reservedness and legal frameworks has resulted in the decision to limit the conditions in which it operates when at level 3 because the back up of an alert driver is, by definition, removed.
Much of this comes down to mindset. Tesla rely on the driver alertness to push the limits, and up to a point that is a rapid way to learn. The somewhat crude analogy is a child learning to walk, an adult is there to catch them should they fall. The downside is Tesla are seemingly putting very little effort into understanding, ahead of time, where they should give up. If your goal is only level 5 then they are betting on the regulators allowing the systems straight in at that level, and matters such as redundancy of cameras, fail safe measures etc etc may all come to the surface (they may not, we don't know). Alternatively they will need to retrospectively go back and learn where the limitations are and learn how to predict and respond appropriately so there are zero driver takeovers on a journey that aren't requested by the car in good time, r where the driver simply wants to turn off the system for other reasons. In Europe, that will be a very hard sell. The mindset here is to start with a limited scope and expand, exactly what the Germans are doing. It is entirely possible we reach Level 4 here with a European based mindset with the likes of Bosch or Mobileye before Tesla, whereas in the US the reverse is true, purely because of the nature of the evidence base and path to implementation.
Examples like stopping because a car is jumping a red light should not be an example of self driving, but of a passive safety system. Drive a competitors car with cross traffic collision avoidance type systems and you see this exact type of feature come into play. The parameters are slightly different, but the logic is the same, the driver should be alerted and potentially prevented from driving, whether manually driving or on autopilot. Tesla should be implementing such measures in all cars as part of passive safety and not limiting it to FSD in my opinion, maybe once the single stack software arrives it will be the case. As a comparison, try driving a modern German car towards a wall below about 10mph when parking, you'll typically find the car will keep braking once it gets within approximately 30cm as a passive safety system.
We've had a "watercooler" chat over city street driving and if and when it arrives in the UK. The material concern is the lengthening of reaction times when a driver has to intervene. Currently, a driver spots a hazard like a child running into the road, and then responds. The distance to stop is a function of how long before the risk is understood, and then how long to physically brake. Modern brakes have dramatically improved the latter, but the concern over things like mobile phones is the former has increased. Now layer of an extra dimension, the car is driving, the driver now has to perceive the risk, and then perceive the car is not responding before they take over. The increase in time could be catastophic. The flip side is the car is looking all the time and always attentive and so will make less mistakes than a human driver. Going back to the earlier point about passive safety systems, there is no logical reason why the car wouldn't intervene should it detect a hazard (just like collision avoidance, but more geared to a city streets environment) but as a passive safety system
In essence, our thoughts on safety is that Level 2 on city streets in the UK will be potentially dangerous because the expectation is the car will react first, and then the driver. If this was passive safety, the expectation is the driver will respond first, and the car will react in parallel.
For our American friends, if you really want a considerably safer driving experience, forget Tesla, simply move to the UK as statistically we already have it on our roads.