A closer look at what just happened with SN112. Pylon Partners acquired the subnet and is turning it into THOR, focused on physical AI and robotics. The problem they’re attacking is massive. Robotics has a data bottleneck. LLMs had the internet to train on. Robots don’t. If you want robots to learn how to manipulate objects, work in warehouses, recover from mistakes, or interact with the physical world, that experience has to be created, collected, labeled and validated. Look at Scale AI. Meta invested $14.3 BILLION for a 49% stake last year, valuing Scale at over $29B And Scale is now pushing heavily into physical AI, building infrastructure to collect, annotate and validate robotics data. This makes THOR very interesting. They’re attacking a similar problem, except using Bittensor incentives to turn some of that centralized production process into an open market. A robotics company needs data. They submit the job and requirements. SN112 distributes the work. Contributors compete to produce it and validators make sure the output meets the customer’s standards. Instead of THOR having to build and manage the entire workforce themselves, Bittensor gives them an incentive layer capable of pulling productive capacity from anywhere. And as robotics scales, the need for real world training data should scale with it. Scale AI has already shown us how valuable AI data infrastructure can become. THOR is basically asking whether Bittensor can build a better way to produce some of it. Pylon didn’t buy a random subnet and slap robotics on it. They picked one of the biggest bottlenecks in physical AI and are applying Bittensor incentives directly to it. I’ll be watching SN112 closely. bittensor:native @THORMachines