[Nevin1901]

I wonder if using a model with a higher TOK/s would yield improvements, as the model will have faster feedback loops