metaverse i‘m unironically getting more bullish on what has seemed to be one of the most useless spending of capital in the last century for a long time - the metaverse. robotics obviously has a deep lack of training data right now, issue is this data is really difficult to get, gopro data & normal image/videos don‘t work. They need telemetry: multimodal, physics-grounded, high-fidelity, temporally aligned data that captures contact forces, joint states, friction, mass distribution, failure cases, and edge conditions. Obviously this data is extremely hard to collect at scale in the real world because it’s slow, dangerous, expensive, and often impossible to even get with sufficient precision. there are now factories being spun up for low level gopro training data to be collected & this data is then sold to labs (example is by @eddybuild) But a rich virtual world where everything has a state vector, everything has a physical model, and everything can be logged down to the tiniest Newton-level interactions that is scalable. A metaverse built on real physics rather than zuckerberg type cartoon graphics is essentially a training data factory. And suddenly those billions of dollars poured into real-time rendering engines, photorealism pipelines, spatial mapping, haptics, mesh compression, motion capture, and high-bandwidth networking start to look much less like a dead end and much more like the early scaffolding for the “unreal world” where autonomous robots will actually learn. metaverse was never a product. It’s an accidental precursor to the simulation substrate that the robotics revolution is going to require. think this is one of the startup categories that remains being interesting and is still in a relative vacuum in comparison to the potential demand. @StreetFDN Batch 03?