Train your model with your model

It's obvious in software to build a mvp then iterate. That's now possible in ML with model assisted labeling.