Overall, I think the rogue looks pretty good. I wish we could see some more of the trickster rogue build powers, because I am curious how Bluff, Intimidate, and some of these other skills might be used in powers, and how class powers might have lots of interesting non-combat applications, but I guess I must make due with what I have.
The limited choice of weapons is a bit surprising, but aside from the cudgel or sap, I think it captures the core flavor of the rogue nicely -- these are all smaller, concealable weapons. The rapier and shortbow are notable omissions.
I like the powers! 3.5E started to attack this with feats based around the rogue's sneak attack (like Ambush feats from Complete Scoundrel), but I think divorcing it from sneak attack is the way to go.
Sneak attack -- no mention of conditions that prevent its use! Immune to crit creatures is the obvious example, but my parties ran into the problem with concealment as well. The damage isn't as ridiculous now, so it doesn't need the same level of restrictions, probably.
Some thoughts:
1. The choice of two Rogue Tactics (Brutal Scoundrel and Artful Dodger) and their associated two builds seems rather limiting. But it does open up space (in the Martial Power supplement, for example) for something like the witty rogue build and Rapier-Sharp Wits as a new Rogue Tactic, and then a slew of powers that interact with this new, Int-based ability.
2. Related to my 1st point, but perhaps one of the trade-offs in trying to make the new system work, and to avoid obvious exploits and problems off the bat, is that they have to narrow the initial set of options a bit. Essentially, go for quality over quantity. Could they achieve both quality AND quantity? Hard to say -- my experience in non-RPG things says this is difficult regardless of the discipline, but I am sure there are counter examples.
3. The only non-combat, non-trap related abilities the 3E rogue received was its skill list. Everything else--sneak attack, evasion, uncanny dodge, trapfinding, trap sense--pertain to combat and traps. So this approach to the rogue is not new; of course, it doesn't mean they couldn't have tried something new. But I think what we are seeing is one of the core design philosophies of 4E: balance each class in combat, and balance each class outside of combat, but don't balance these aspects with each other. Balanced combat options requires many different abilities, powers, etc., since there are so many facets to a fight, and because the way a wizard operates should be different that a fighter or rogue. Balancing outside of combat -- could it be they are achieving balance and relevancy for every class by primarily trying to tune the skill system? Already the ability mod + 1/2 level math makes it so the difference between a trained and untrained character is smaller than in 3E (where, at 1st level, the diff. could be easily be +5, and could exceed +20 at high levels).
After all, when people talk about non-combat abilities, it typically boils down to the skills they choose, and perhaps a few feats that interact with them. If each class has a skill space where it is immediately useful, and then through feats can choose other skills to fill out the character and do things that are non-canonical for the class (like a fighter with Diplomacy, for example), I think this is a good thing.
4. Skill choice: Hmm, the rogue gets two trained skills for free, then gets to choose 4 more from a list of 8 others. Seems limiting. But math says there are 70 unique combinations of 4 skills selected from a list of 8. That seems like a fair number of options -- certainly less than 3E, but that largely comes from consolidating the skill list. Certain combos probably make more sense than others. The only complaint I see here is that all rogues must be stealthy (but this seems a bit silly -- this is the D&D thief/rogue to a tee) and good at picking pockets, disarming traps, and opening locks (this objection I understand). But this is probably just like complaining about every rogue getting sneak attack -- each class is going to have defining features. It is the limitation of a class-based system.