[nl]

I agree about it being early days.

Reformer is a good example that I'd missed.

Neural Tangents is another paper demoing a framework.