You can see that the number of new active infections has become more linear at places like the worldometer. I’ve been fitting this data with a lumped parameter SIR model (The key difference between a standard model and this one is that I model untested or asymptomatic people as a seperate category than infected.) No single model fits the newest data. The difference between the yellow ‘I’ and blue ‘I’ (infected and tested positive active cases) in the graph above is simply lowering the apparent infectivity of the carrier category (untested or asymptomtic).
The difference doesn’t look big in the log plot but it’s 100% of the value and significant. (The errors in the old data are likely due to poor testing or assignment rates.) It’s just beginning to show up in the death rates. This is really encouraging.