IIRC the rate of publishing was superlinear thus the curve of actual publications goes faster than the quadratic function.