My initial response was very negative; it's been tempered a bit in places upon further reading, though.
The changes to the fighter? Fine. Great. I miss Jab and Snap Shot, Whirlwind Attack seems to make Cleave obsolete, and Parry should be fighter-specific. Tumbling Dodge is kind of cool. The extra attack at level 6 seems a bit clunky at first, but I can see it adding a lot of cool options... I'd almost rather see the action points from the first playtest return instead, though. (And I certainly didn't expect them to cut out all the good fighter specialties instead of adding more.)
The cleric? I had to laugh when I saw we're basically back to the 3e cleric, with fewer spells and domains. And the wizard traditions are quite underwhelming. I think taking away the unlimited cantrips/orisons was a terrible idea, and not just because it ruined my favorite feats. It adds unnecessary complication and only makes every wizard in a given tradition rely on the exact same spells over and over. (How many battle wizards do you think WON'T be casting Thunderwave every single combat?) I think it would be a much better idea to have a narrower list of 0-level spells to choose from and then have each tradition/domain grant an additional 0-level spell not on the general list.
As for the spell progression... pure silliness IMHO. A single spell at first level? Were people really clamoring for that? It seems very likely they just wanted the numbers to work out so that a level 10 character has about 10 spells. Okay, here's another way of achieving that:
Code:
Lvl 1 2 3 4 5
1 3 - - - -
2 4 - - - -
3 4 2 - - -
4 4 4 - - -
5 2 4 2 - -
6 - 4 4 - -
7 - 2 4 2 -
8 - - 4 4 -
9 - - 2 4 2
10 - - - 4 4
See that? You've got 8 spells per day by 4th level and it stays that way indefinitely. If you want a lower-level spell you can prep it in a higher slot. If this "overpowers" spellcasters then the spells themselves should be toned down; no way a dedicated spellcaster should be expected to go entire battles without casting a single spell, like clerics are up to about level 5.
The rogue? Bleh. 9 of its 14 maneuvers are shared with fighters. Sneak Attack is just a conditional Deadly Strike. This class is basically crappier than a fighter in every way except that he has more skills (and is better with them). I know everyone likes to be sarcastic about how the fighter should be best at fighting, but this is a frankly boring way of achieving that.
Here's a quick "fix": give the rogue an entirely different scale for expertise dice. Give him 1d6 per level. Return Sneak Attack to its previous iteration (only when you have advantage). Add a Flanking Strike maneuver that lets you add only the best expertise die result to damage. (Take away Parry too.) That way rogues can still do crazy damage with SA, and when they use dice for other stuff they get more reliable but lower results. (More but smaller dice.)
I'd still rather they keep expertise dice for fighters, but at the very least they should work to make the classes more unique in the way they use them.
Overall, I actually like a lot of the specific changes to spells and so on, but I'm kind of shocked by some of the changes they made to the classes. Mearls made a big deal about how universally popular the at-will cantrips are; why would they screw that up to satisfy a few pre-4e purists who STILL can't make a damn wizard without unlimited cantrips? Do they honestly think that players are going to be excited to roll up a cleric that can only cast one spell a day? After thousands of forum posts debating whether and how other martial classes could use expertise dice and still remain unique, did they really think giving rogues five unique maneuvers and crappier HP and armor proficiencies would be a step up in class design?
I hope that jrowland is right and they're shaking things up to see what rattles.