Here’s the explanation at Covid.us.org: The Covid-19 Death Rate Has Dropped by 2/3rds. The comparison is weekly total of new cases to weekly total of new deaths. The new deaths is from a week 14 days later than the new cases week. That’s because it takes about 14 days or so from diagnosis for someone to die of Covid, if they are going to die. And that timing is of course an average.
If you were diagnosed with Covid in the week of April 5 to 11th (Sun. to Sat.), your chances of dying were about 6.88%. This uses the week beginning April 19th (14 days later) for the weekly new deaths data.
But if you were diagnosed in the week beginning June 14th, your risk of death was 2.21%. This uses the most recent full week, beginning June 28th, for the weekly new deaths data.
That is a difference of 3.11 times. So the risk in late June/early July was less than a third of what it was in April. And that is just over 2/3rds less risk of death.
The article here discuss why that is. And it is not all good news.