Let's start with New York City. A few weeks ago, NY State announced that there were some disturbing increases in COVID cases. Hot spot zip codes were identified (see this announcement).
Contrary to what some believed, these hot spots were *not* places that were hit hard in the Spring. Of the 13 top hot spot zip codes listed that are in NYC, only 1 had COVID cases adjusted for population that were among the top 10 worst COVID zip codes in NYC (out of 179 zip codes for which there is data). This is (sort of) good news, and indicates that the worst areas may have developed, dare I say it, something close to herd immunity. In order to guess the percentage of infections by zip, I took the number of confirmed cases and multiplied by 10. This very rough estimate often corresponds with another rough estimate: deaths divided by 0.6% (assumes a 0.6% death rate), with a few exceptions.
Using the confirmed cases per capita, adjusting for confirmed to non-confirmed, the worst 10 zip codes from the spring all had cases in excess of 40% of the population (with the highest at about 50 % -- Corona queens was at 47% using the case measure and 67% using the death measure). The current hot spots, on the other hand, averaged 33%. This is only a little in excess of the city average of 28%, using the same rough measure. The other good news is that after peaking a couple of weeks ago, NYC is coming down again, indicating that at the moment, we do not have an exponentially increasing spread (so herd immunity for the moment, perhaps).
How about the rest of the US? The hot spots, at least on a state level, are those that fared best from March through the summer. The five states with the most cases in the past week (population adjusted) are North and South Dakota, Wisconsin, Nebraska, and Montana (see here). These states' ranking are among the lowest in deaths per capita so far in the pandemic: 22, 32, 41, 42, and 43 (out of 50 states and DC). In other words, in some sense, this is the tail end of the first wave rather than a second wave. Furthermore, at least according to the NY Times, the worst outbreaks are now in rural areas whereas the worst outbreaks in the spring were in cities.
The good news of the summer was that COVID does not spread well outdoors (we know this from studies like this https://www.medrxiv.org/content/10.1101/2020.02.28.20029272v2 and from observational data like protests and Trump rallies, which I mention because they are generally mask-free).
But now we soon need to face the bad news of the winter: COVID spreads very well indoors. There seems to be little doubt that as people gather indoors again, cases will rise, because we are not yet at or particularly close to herd immunity nationwide (see below). However, there is good reason to believe that the winter won't be like the worst of the Spring, because we are *part* of the way down the road to herd immunity and because COVID is much less deadly than it was at first. The lower death rate is partly because the most susceptible have already died, partly because we have found better treatments, partly because the susceptible are being more careful than the young (this changing the profile of those sick toward those who do not get very sick). Even in states that are lifting all restrictions, many people will continue to take precautions against COVID. This is good news for what looks to be a long winter with no vaccine until spring.
Where are we in terms of infections and herd immunity? If we assumed only one-tenth of cases have been confirmed, there have been nearly 90 million cases in the US. If we assume about a 0.25% death rate (about half that of NY) nationwide, then we have had about 45 million cases. That means between 15 and 30% of the population has gotten COVID (see this article for estimates as of July, which put the nationwide rate at around 10% and the Northeast at around 30%, but note there have been huge increases outside the Northeast since then). In any case, even these rough estimates make it clear that, assuming herd immunity is in the 40-50% range, we have a ways to go in most of the country.
There are probably places, like parts of the northeast, where we are close enough to herd immunity that the winter outbreak will be quite mild. However, looking at NYC zip code by zip code, it is clear that rates are very localized, so there are some areas, like where I live, that the virus barely touched. If people in these places starting having large indoor gatherings (which they wont), case would explode here.
Nationwide, are we at least at the point where cases and deaths begin declining? I think the answer is probably yes. Why? As I said above, the record-breaking cases we've had the last couple of days nationwide are likely the tail end of the first wave. The second wave will almost definitely be milder, except for places that really had no first wave.
What does this mean in terms of numbers of people dying? Given that most of the country has had a first wave at this point, it seems very unlikely that last month's prediction of 400,000 deaths by year's end will be correct. That prediction would mean at least 40 million more infections in the next 2 months, and, in particular, it would mean probably twice as many infections in the next couple of months as we have had up until now (the death rate has dropped to half or less of what it was, so to have the same number of deaths, you need at least twice as many infections). You also need to consider that, even in places with no government mandates or closures, a large majority of people continue to take some precautions against the virus.
So, yes, there will be a second wave. But it will be nothing like the dire predictions we are hearing, and it will be less severe than the first by a good margin.