We are multiple orders of magnitude away from Landauer limits - so next big thing in matmul could be photonic multipliers - there’s a bunch of them coming up in the next 3? years. So that’s a 2-4 order of magnitude improvement. Sigmoid?