First, set up a Conflict Web. Start by setting up your factions that are involved, and why they are competing/conflicting. This is more to give you a set of motivations for any given group, leaders, etc. and allow you to simply improvise based on the group's needs/ambitions.

The Conflict Web is not static, it's a starting point. So you may easily see characters shift alliances or make temporary truces to accomplish goals.

Second, once you situate the PCs into the scenario, look at their goals, and likely problems they will face in terms of Logistics and Politics. This is effectively similar to how Apocalypse World produces "Fronts".

After each session, look at what the PCs attempted, who was affected, whether any NPC groups made major moves and figure out who is going to react and how. You can choose to update either the Conflict Web or the Logistics & Politics list, though I usually find myself only having to do serious updates after 3-6 sessions because it's relatively easy to track what happened with simple notes.

Both of these tools can scale up or down, so you can do intergalactic empire politics or the 28 guys stuck in a prison together, based on whatever fits your campaign.

Dungeon World is a narrative game, at its core, that distinguishes itself from D&D in the way it tells stories. The innovations are in the core philosophies and mechanics. Let me address each of your points in turn:

Moves as Powers

Moves are NOT just powers. Many are closer to D&D's feats. Others have no mechanical effect at all. Some simply tell you that your character can do a thing, without giving any rules for it. The real key to class moves is that they provide distinct flavour to the characters, as well as story-telling hooks. Take a look at the paladin's Quest, or the fighter's Signature Weapon, and notice that they're more about interesting character details than they are mechanical effects.

No Scaling

Love this. A level one character and a level 10 character can coexist without any real issues, and being level 10 doesn't mean that early monsters are no longer a threat. DW's implementation of this isn't anything particularly special, but it works really well in the context of the rest of the game. Because stats don't increase by that much over the course of a campaign, you still end up with interesting results for die rolls (see Interesting Failure, below).

No Initiative

I also love this, but it's a tricky beast that takes a lot of getting used to. The GM should be directing the action and the spotlight such that every player gets a chance to be involved. But note that there are certain situations where it makes perfect sense to focus on one character or another. Combat scenes in DW should follow the flow of the action, not be restricted to a turn-by-turn basis.

Less Tactics

Of course there's flanking. But it's handled narratively, not mechanically. See, for example, the thief, who gets to use Backstab on a surprised enemy. One of the ways to accomplish such surprise is for the thief to have positioned himself behind the enemy while another character distracts him. Or perhaps a bard chooses to taunt and distract an enemy, and meanwhile the fighter attacks the enemy from behind. Remember that if you attack something that isn't fighting back, you don't make the Hack & Slash move, you simply do damage (or whatever else you're trying to do).

Overall, the players have infinite freedom in what they can try during a fight. The key is to respect the fiction and pay attention to the situation. If there's interesting terrain, tactics are involved. If the enemy has only one weak point, tactics are involved. If there are many enemies, tactics are involved.

XP for Misses

Not earth-shattering, no, but this mechanic encourages players to try things they might not normally do, which is no small feat. Rewarding players for failure makes the players more invested, and even excited about failing. More on this in a second.

There are plenty of the innovations that you've missed entirely in your analysis. These tend to appear in the core of the rules, not necessarily readily apparent. They're apparent in the philosophy of the rules, if not the mechanics themselves. And many of them are in the GM-facing rules, not the player-facing rules.

Interesting Failure

Failure in DW is interesting. Note that all moves involving dice rolls have only three possible results: success, success with complication, or failure.

when you fail, the GM has every right (and in fact, is mandated by the rules) to drop a hard move on you. A failure is never simply "nothing happens." Instead, a failure leads to a change in the situation: you take harm (and not necessarily just HP damage), you get separated from your friends, you lose your stuff, you encounter some portent of future badness, etc. This is the real key, here, that keeps the game moving and keeps the stories interesting.

On that note, the most common result is success with complications, which is also interesting. Take a look at Defy Danger, probably the most commonly used move in the game, in which most successes come with a worse outcome, a hard bargain, or an ugly choice. Plain old success (e.g. you dodge out of the way of the falling boulder) isn't as interesting as complicated success (e.g. you dodge out of the way, but you either lose your weapon or your backpack in the landslide).

Mechanics that Encourage Storytelling

Take a look at the Spout Lore move. When a player request information of the GM, the GM is encouraged to ask the player how they know that information. The player gets to use that as an excuse to add information about their character (and the world) to the story. Or how about the Bard's A Port in the Storm? It mandates the GM tell you something new and interesting about the world every time you set foot in a place you've been before. Moves like these litter the rules.

Collaborative Character Creation

Players create characters during the first session, with each other and with the GM. The GM is expected (again, mandated by the rules) to ask questions, using the player's character choices to inform the setting. In the last game I ran, one player decided he wanted his Fighter to have scarred skin (as is one of the look choices on the Fighter's character sheet). I asked him how he got those scars, and he decided they were from fighting in a war. I used that idea to decide that the world was one that had only recently found peace, with tensions still running high while people slowly reconstruct civilization. Had the player suggested that he got the scars fighting as a gladiator, for example, that would have let to an entirely different setting decision. And that was just one question out of dozens that I asked the players.

And then there are bonds, which are statements of relationship between characters. This gets players thinking not only about themselves, but about how they relate to each other. The bonds themselves are vague enough to leave a lot of room for the players to come up with specifics, but defined enough to really cement the relationships.

More

There's probably a bunch more that I'm missing, but I've written enough, methinks, so I'll leave it at that. The other answers here are also really good.

It should be noted that most of these things aren't revolutionary in and of themselves. Each has been done before in other games, but what makes Dungeon World such a great game is how well all of it is put together, and how cleanly each mechanic is implemented.