This is probably a dumb question, but I am going to ask it anyway. In light of open source models getting so popular, the chance any of us have to train a large frontier sized model on our own, at least for most of us, just isn't possible. But what if we created a skill, plug-in, some sort of tool,
