That article neglects that the plants built by the end of the program were not the same as the plants built in the beginning. Within each "pallier", a speedup can clearly be observed. Likewise, costs can hardly be compared if you don't look at the same models.
I have no idea how a speedup appears to you. On the graph it is quite clear that, with time (and therefore experience), the average amount of time needed to build one is raising.
That's 2-3rds of the builds showing a speedup. The results are even more striking if you calculate the correlation between start date and build time for any 2 reactors of the same model.
Yes, I considered the long-term experience gain (columns, not lines: from 65 to 151 and from 60 to 104).
The first nuclear plants were theoretically the most difficult to build as the local industry was less adjusted to building such things, especially specific components.
As those reactor 'models' were very similar (there is no quantum leap) pertinent experience (processes, tooling...) accumulated.
However there was no reduction of 'intensity' (investments, amount of simultaneous building projects, foreseeable projects...) as all this was encompassed by a huge national programme (the 'Plan Messmer').
Therefore it seems that both min and max time to completion should diminish with time.
> Yes, I considered the long-term experience gain (columns, not lines: from 65 to 151 and from 60 to 104). [...] As those reactor 'models' were very similar [...]
They're not the same buildings. A N4 is much larger than a CP0, uses different technologies, has more safety features, produces much more power, etc.
To compare with another tech topic, that's like expecting SpaceX to design or build their spaceship faster or cheaper than they designed their falcon. That's unlikely, even though falcon knowledge definitely benefited the design of their new craft.
Not exactly the same but same generation, architecture and design (Westinghouse), slightly (not fundamentally) enhanced. Stating that new features add such a large amount of work (relatively to the total amount) that it compensates for the knowledge gained thanks to previous projects is debatable.
Between the oldest (CP0) and newest (N4) aren't the key differences limited to a same machine and command room shared (CP0) or not (N4) between reactors, scale (CP0's nominal power being lower), and details related to fuel rods and pipes? In which way are they dissimilar to the point of absorbing the effect of gained knowledge and adding such delays?
Even the shiny new EPR is a mere enhancement of the core design dating back ~1970.
Sorry, I don't know enough about SpaceX to have an opinion.
Negative learning by doing is very easy to explain: as you go, you realise many perils and dangers of this thing that you were not aware of in the beginning, and those result in a lot of safety precautions added in subsequent projects, increasing the costs.
I begin to understand the logic behind the US derailings.
The idea here is that maintenance is much, much cheaper than rescue operations. If 100 more pumps let you run for years without a scram, go along and order them.
[1]: https://fr.wikipedia.org/wiki/Centrale_nucl%C3%A9aire_en_Fra...