The way Archer seems to generate so much pace through such an easy action really amazes me. All the express bowlers that I've seen over the years either had super long run ups (Lee,Bond) or extreme slingshot actions (Tait) or both (Shoaib). Archer just seems to amble in and roll his arm over and bang 95 miles. That being said there are obviously areas in which he should improve like pitching the ball a yard fuller. The way he keeps beating the bat on the shortish length is awesome and all but pitch it a little fuller and snag the edge. We really dont want another Morne Morkel do we
With respect to pace, according to CricViz (yeah, I know...) Archer's average was 140.9 over the course of the test, while Starc's average pace in tests this year is 145.53, Bumrah's has been 141.15, Wood's 140.47 and Cummins' 140.18. The difference between him and the others is that he is extremely variable, moreso than any other bowler I can recall (although Duanne Olivier is not too far off). I remember at least one instance where a bunch of 136-140 kmh deliveries sandwiched a 150 kph one. I can't think of another fast bowler who can send down a random ball 10 km/h faster than the previous.
I suspect that this is what makes it hard to play him, unlike what other people have said I don't think that the batsmen actually have much trouble picking up his length and he almost seems to telegraph his head-high bouncer, instead the short of a length skiddy ones are the ones that hit people, especially on a surface with some variable bounce. It's the pace variations that are hard to pick, and the relative slowness of his runup will also fool a lot of batsmen (although this can be gotten used to and was once more common than it now is).
With respect to the mechanics of generating pace, he hyperextends* a large amount, though I think not quite so large as Akhtar, while having longer arms than him, so it's not all that surprising that he can generate a lot of pace. Akhtar is actually an interesting case for a long, fast run, because his action was somewhat 'knee dominant' (back foot pointing backwards, back leg bowing type action, generate power from quads) which means he wasn't actually getting full value out of that run up. Archer is a conventional 'hip dominant' bowler (quick off back foot, straight front knee, generate power from glutes) who are usually better at transferring they momentum into pace, whereas slingers rely on strength and elasticity. The difference between Archer and, say, Wood, is the hyperextension and the fact that he straightens the hyperextension through delivery, in effect 'throwing' without being legally considered throwing.
Without that his 'base' pace (136-142) would probably his ceiling pace unless he adopted a more forceful action and faster run. Variations in the hyperextension are almost certainly the reason why his pace varies so much, and although no expert I would venture to suggest that it is why his short balls are faster than his fuller ones and he seems more comfortable bowling them. This because the straightening of a hyperextension is somewhat uncontrolled, and will both be more complete for a shorter ball delivered later in the swing, hence faster, and where there is that lack of mechanical control aiming at a closer spot may be easier with less error in line. I think it may give him some Charlie Griffith style unpredictability too.
*Although where extension (i.e. throwing) becomes hyperextension (which is considered legal) is hard to discern without fairly technical testing and an umpire would be justified in reporting him if they wanted to, Akhtar hyperextended if anything more markedly so he'll almost certainly be fine. I have seen at least one study instance which implied a bowler or bowlers being tested were doing both, which is interesting.