I’m looking at a mountain range projected on a wooden table. The mountain range isn’t part of a flashy game or art project. I can’t reach out and touch it like a real object. Thanks to some still-in-progress software optimization, it glitches a little when I move. And I’ll never purchase the high-end augmented reality headset that’s creating the illusion — the Magic Leap 2, set for launch later this year. But the scene is remarkable for an important reason: I can actually see all of it at once.

Magic Leap was once known for its theatrics and huge promises, but the massively funded yet embattled startup has spent years trying to get back to Earth. It laid off a huge portion of its workforce and changed CEOs in 2020, scrapping its mass-market AR plans to focus on healthcare, manufacturing, and defense. The Magic Leap 2, formally announced in 2019, is supposed to cement its presence in those industries. In reality, the company’s future still seems uncertain. But based on a limited demo of a version with complete hardware and in-development software, it’s launching a genuinely improved second-generation device including a markedly better field of view — taking a step toward assuaging one of AR’s enduring pain points.

Like the 2018 Magic Leap 1, the Magic Leap 2 includes a pair of dark gray goggles wired to a puck-like computer that you can hang from a shoulder strap or clip on a belt. Those goggles refract light from small LCOS displays through multilayered lenses that project holographic images into your surroundings. But they’re doing it in a much trimmer package. The Magic Leap 2 weighs 248 grams to the original’s 316 grams, which was already svelte compared to the 566-gram Microsoft HoloLens 2. Between the weight reduction and an optional over-the-head strap, it fit me more easily and firmly than almost any other smart glasses I’ve tried — albeit for a roughly 30-minute demo, which is far from the full workday Magic Leap says it’s designed for.

You won’t find a dramatic design overhaul in the Magic Leap 2’s simple motion-control remote, but internally, the company has made a significant change. The Magic Leap 1 tracked its controllers’ movements with electromagnetic fields. But, citing problems using magnetic sensors around some industrial equipment, Magic Leap has switched to optical tracking that incorporates both headset-based sensors and cameras mounted in the actual controller. The remote isn’t built to give you full-fledged virtual hands the way many VR controllers do, and I didn’t get to try any complex object manipulation, but it certainly feels functional enough for simple point-and-click interfaces. (I checked out an extremely unofficial Magic Leap Beat Saber clone, but it’s hard to judge hardware performance from a rough prototype app.)

This is all potentially great for Magic Leap’s enterprise customers, around 35 of which are testing the Magic Leap 2 ahead of a release scheduled for the third quarter of 2022. But it doesn’t mitigate the device’s long-running tradeoffs. Offloading electronics to the puck makes the headset lighter and more comfortable than the self-contained HoloLens, for instance, but it means you’re walking around with a long wire attached to an odd-looking computer. While Magic Leap CEO Peggy Johnson says this hasn’t been a major issue for current customers, who use the device for things like training simulations and medical diagnoses, it indicates the limitations that high-end AR headsets still face.

In an evenly lit and monochromatic demo room, the Magic Leap 2’s holograms look great by current AR standards. (That means the images are still a bit transparent, but they’re crisp and vivid, and text is easy to read.) A new feature can also make them stand out against the real world by dimming parts of your vision to near-darkness — bright lights still poked through, but I had to strain to identify other objects while using it. And while I didn’t check out any sophisticated blends of real and virtual space, objects stayed pinned to one place in a way consumer headsets like the Nreal Light can’t manage. That makes sense, of course, given the Magic Leap 2’s far higher price point: it’s supposed to cost slightly more than the Magic Leap 1, which starts at $2,295.

On the flip side, both generations of Magic Leap headset — like some other AR devices — use dark lenses that consistently dim your vision and make you look like you’re wearing giant sunglasses indoors. This may not change any time soon, says CEO Peggy Johnson, although Magic Leap is experimenting with solutions. “That’s certainly a piece of feedback that we’ve heard about this — that you want to be able to see others’ eyes when they have it on,” she says. “There’s no doubt that over time that will come. I couldn’t really guess when that would be.”

Magic Leap no longer promises some secret sauce that’s supposed to make its headset completely different from the competition. As VR/AR expert Karl Guttag has noted, the Magic Leap 2 scraps a system meant to simulate multiple focal distances, one of its few really unique selling points. (Johnson says this was a worthwhile tradeoff to reduce the glasses’ size, although she doesn’t rule out the possibility of bringing it back.) But the headset seems to deliver its most exciting promised feature: a more natural-feeling field of view.

The Magic Leap 2’s specs have largely been revealed already, including its field of view, which offers 70 diagonal degrees compared to 50 degrees on its predecessor. That’s a fraction of humans’ natural field of vision, and it’s smaller than the VR headset standard of around 110 degrees. Even so, the Magic Leap 2’s field of view feels less immediately off than the first Magic Leap or the comparable HoloLens 2 — where holograms can get abruptly truncated with small head movements and leave the impression of looking through an invisible-framed window.

That window still exists on the Magic Leap 2. But it’s gotten a lot taller, making it less likely that a virtual object will exceed its size. Meanwhile, the headset’s goggle-like rims obscure some of your peripheral vision, creating the sense that there’s a physical object and not a digital limitation blocking your view.

Like practically every AR company, Magic Leap envisions making a pair of self-contained, normal-looking glasses for a mass market — Johnson even speculated about the possibility of AR contact lenses. But she downplays the possibility of releasing one any time soon. There’s also persistent speculation that Magic Leap’s long-term goal is getting acquired by another company. Addressing the possibility of an acquisition, Johnson says Magic Leap has been focusing solely on the launch of the Magic Leap 2. But “I do think a successful launch opens up all sorts of opportunities,” she says.

If Magic Leap isn’t acquired, it will have to build a sustainable business with its comparatively narrow slice of customers. Citing some optimistic projections about AR adoption, Johnson thinks it’s a viable short-term goal for Magic Leap — “I think there’s enough interest in all those three areas where any one of them could actually be a business in and of themselves,” she says. Beyond that, it’s taking a leap of faith.