I was slightly unclear. Your statement is true.

I do not say that a modern nuclear reactor can level a city. I don't even claim or disclaim that the worst currently running nuclear reactor can level a city under reasonably imaginable coditions (I tend to agree that the fallout will be a problem, but a full-scale nuclear explosion is very unlikely but I have not enough evidence and knowledge to be sure either way).

I describe a situtation of the research of nuclear fission. Imagine that someone knows that a bigger pile of uranium emits more radiation and wants to build a power plant based on this in 10–20 years. Some research is done to be able to predict the behaviour of such a system — of course, there are no power plant designs from Earth-2010-our-timeline.

How should one do the research to prevent Chernobyl type disasters, minimize the risk of Fukushima type disasters and not find something that makes military build a nuclear bomb before first nuclear power plant is built?

Note that one needs to do enrichment both for a power plant and for a bomb.

It is true that simply piling even warhead-grade enriched uranium will not lead to a weapon-scale explosion, but the results of building a reactor without careful research into implications are not likely to be good.