A camp in Model Interpretability that says we should reverse engineers model completely. Understand every layer and parameter