Lots of excitement last week about finally understanding neural networks, thanks to @nickcammarata's essay.

Nick revisited early research he helped produce and found patterns he had missed.
And the reaction was: oh yes, neural networks are clean, we can see through them now.

There are indeed some fascinating findings. But is it all so clear?

As always, I want to see the whole picture. @nabla_theta's response asks whether these explanations can survive tougher tests.

Here are both sides, and what we still need to investigate: https://www.youtube.com/watch?v=VYzjtj-1CNk