The design of this system, from what I've heard so far sounds
insane.
For some background, systems intended to automatically override pilot input upon detecting an impending stall have been used on large, complex aircraft for a long time. The most common mechanism is a stick pusher, which mechanically pushes forward on the stick to mimic what a pilot should do about an impending stall. The force of a typical stick pusher is low enough that a pilot can overcome it with arm strength.
A stick pusher, or a pilot pushing forward on the stick is not enough for the aerodynamics of the 737 MAX in certain circumstances. The position of the engines means the thrust tends to push the nose up, and at low speed, the elevator controlled by the stick alone is not enough to push the nose down when nearing a stall. Instead, the trim must be used, which moves the whole horizontal stabilizer rather than just the elevator. That's not part of the procedure on most aircraft, including other 737s, so regulators would not allow pilots rated for 737s to easily transfer their rating to the new aircraft, if they were willing to certify such an aircraft at all.
MCAS is the solution to this problem. It automatically uses the trim to force the nose down with greater authority than the elevator alone could provide. While the concept is sound, the implementation strikes me as reckless.
First, the system uses only one of the two angle of attack sensors to determine that the aircraft is nearing a stall. Such sensors have a fairly high rate of failure relative to the safe operation of a commercial aircraft. As far as I can tell, causing the system to use both sensors is a software change and all models of the aircraft had the required hardware. I can't imagine a good reason the system was designed this way, especially since Boeing has included MCAS on a previous aircraft, the tanker version of the 767, and it did use two sensors in that application.
Second, activating the system automatically, and allowing it to contradict the pilot's attempt to pitch up using the elevator seems like a mistake. While stalls are almost always the result of pilot error, I'm inclined to think automated systems should let humans have the last word. A better design might have been to automatically trim nose-down in response to full forward pressure on the stick at low speed. That could be in combination with a stick pusher providing full forward pressure on the stick when an impending stall is detected in the absence of pilot input to the contrary.