Pretty big error bars on any estimations, with the low number of samples. But yeah, I wonder if anyone has tried to give a confidence interval for "tech level over time".