So, let's look at what we expect to be able to buy given what we know. It has become clear that whatever secret sauce they have is far less exciting than was guessed at. That does not mean they don't have something up their sleeve it just means it might not be as much of a leap forward as speculated.

No big jumps in resolution

This was predicated on a novel display technology, the fiber scanning display, that does not appear to be ready. A more traditional 1080p display is likely what we will get. The quote from reporters "There are no pixels" will not be true for the first product.

No hugely increased field of view

Without the high resolution, Magic Leap will have a hard time increasing field of view by a huge margin. They might have something interesting with regards to optics but the further you spread out the pixels the worse the image will look. Expecting something better than hololens but far worse than the Oculus or Vive seems reasonable.

Low fidelity variable optical focus

This is core to Magic Leaps philosophy. They need something here. But there has been doubt cast on how well this will work. Some speculate there will only be two planes of focus. Others think it will all be done via eye tracking and software. This area is likely the most ripe for a surprise. They might have something interesting. The "photonic chip" is the only part of their product they have shown publicly so they are clearly proud of it. We should set our expectations low though as this is a hard problem particularly as the display technology they hoped to develop fell through.

Inside out tracking on par or a bit better than tango

If there is one thing HoloLens should be proud of, it is inside out tracking. It does a great job at this. It is the best there is. We can hope magic leap is able to build something similar to this but likely they will fall sightly short. This isn't that damning though as you don't need perfect tracking to make a useful product and it is something that can be improved as HoloLens has proved. I suspect we might get something similar to what google has achieved with project tango. That is to say, we don't get perfect tracking but it is passable. Objects appear relatively stationary with only some jitter.