We saw this happen with Stable Diffusion and it's not surprising we see this happening here. There is a lot of interest in taking these models that are in striking distance (single order of magnitude) from running inference and training on consumer level hardware and as such a lot of energy is going into making the optimizations that can get us there.
Generally speaking, research is not usually done with consumer usage in mind, so what this is, and Dreambooth etc. for Stable Diffusion was, is that gap between researcher software and accessible software being bridged.
So many parts across the stack need to work well for this to go well. Early support for popular software is a good example. This goes from partnerships all the way down to hardware designers.
I'd argue it's not about engineering more than it is about good organizational structure.
The bangle.js:
- ships significantly faster
- has an always-on display with similar 4 week maximum battery life
- can be updated without flashing
- has a thriving app ecosystem
It must be trivial to use one of these approaches to generate a removable background and render the background with something else, like a static image, video, or 3d software
A GPU from 2008 has to fill a quarter of the pixels at half of the refresh rate and has no latency, weight, or heat requirements. The Quest 2 has to carry a massive burden being on a face and fill nearly 4k pixels at 72 FPS. Meta has stringent requirements on maintaining that FPS to prevent VR sickness
Classic problem of the game is too easy for the developer because they've been playing it for so long. The controls are unintuitive but fun if you're willing to take a beating for a while. Even after a long time learning tho there's a strong lack of control, a little bit of rotational friction tending to keep the ship upright would go a long way