The existing "big" nuclear reactors are already more or less in the "happy" range. The problem is irregular building schedule (you can have 10 reactors in works today and 0 for the next 50 years) and extreme regulatory overhead.

Ideally, you need a "pipeline" with one reactor being built every 2-5 years and streamlined regulatory environment where you can use the same project "template" with minimal adjustments. This way building expertise is not lost (i.e. the industry does not need to re-learn everything for every project, which is very costly) and even can be accumulated to reduce costs further. It applies not only to the building part, but also to regulatory bodies.

The problem is that you need to operate ~100 reactors for this, so it can work only for large economies able to provide the "baseline" demand. China and to a lesser extent Russia do this, which does result in significantly lower nuclear plant building costs for them. Europe, SK, and US have dropped the ball hard here.