https://gis.cdc.gov/grasp/COVIDNet/COVID19_5.html
Since we can never know how many people have had COVID-19 unless we tested every single person, we need to find another data source to determine if this is really getting better or worse. Without knowing the actual number of who has it and who doesn't, talking about mortality rates is a useless exercise.
Hospitalizations seem to be a good source to use. People who have severe symptoms would seem to go to the hospital on a consistent basis, regardless of how much testing is or is not being done. It appears that rate is steadily declining. I'll be interested to see the data for Week 26 when it is available.
So the conclusion I draw from this data is one of two things must be happening. Either the virus has mutated into a less virulent strain that doesn't lead to as many hospitalizations, or it is just going away and the increase in testing is leading to the increase in cases. Either result is good in my book, and truthfully my guess is it's a little of both.
As you mentioned, deaths are not a good data source to use because of the delays in reporting. Also, I would assume that most people that die had severe symptoms and would have been admitted to the hospital, and thusly included in the CDC data above. And I believe hospitalizations are reported more timely than deaths.