I agree that there's opportunity there. In fairness though, IBM does have a model (which we laugh at) and they are a leading supercomputer manufacturer. I think part of the problem may be data acquisition for model input. NWS has a large network to gather data that would be really expensive to replicate. I'm not sure that data is all readily accessible. The old data adage "garbage in, garbage out" is very accurate. So even if IBM or Musk could build an epic supercomputer and program top-notch physics into it, the model would still be crap without good data.