The reversals are the striking part. Governments that spent thirty years running down their nuclear capacity are now signing agreements to build it. A reactor shut as uneconomic is being brought back. A second small modular design has been certified, and the announcements have taken on the tone of something settled.
What changed is not the physics and mostly not the safety record. It is that demand forecasts moved, and the alternatives got worse relative to a thing whose disadvantages were already known. That is a legitimate reason to change position, and it deserves better than either triumph or suspicion.
Lee Kuan Yew was unusually willing to drop a position when the numbers moved, and he attached one question to that willingness which is worth borrowing now, particularly by the people currently most enthusiastic.
Has it been made to work, by anyone, near enough to here?
His test was narrow. An idea earns a place by producing a measurable result somewhere, under conditions close enough to yours to be informative — and the doctrine it belongs to goes the moment the result says it should. What he refused was the argument from principle, on either side: that a thing must work because the theory is elegant, or must fail because of what its supporters are like.
The reason this matters right now is that the word nuclear is currently covering several propositions with very different evidence behind them, and public argument runs them together.
Restarting a unit that operated for decades, with an existing site, a trained workforce nearby and a known performance history, is one proposition. The uncertainties are real but they are engineering and regulatory uncertainties about a machine that has demonstrably run.
A certified design with no unit in commercial operation is a different proposition. Certification is a statement that a regulator has examined a paper design and found it acceptable. It is a genuine achievement and it is not an operating reference. Between the two sit the first-of-a-kind cost overruns that emptied the sector in the first place, and those overruns did not come from bad physics. They came from building something nobody had built before, which is what a first unit is by definition.
A signed deployment agreement is a third thing, and it is the weakest evidence of the three. It is a plan.
So the practical form of his question, for any of these, is: name the unit, somewhere in the world, that is currently generating under conditions close to the ones you are proposing. If the answer exists, the argument is about cost and schedule, which is a normal argument. If the answer is that the first one is under construction, the honest description is a bet, and it may still be a good one.
Watch out for
His test has a failure that is more relevant here than anywhere else, and it is in his own record.
Whoever holds the power decides what counts as working, and the criterion quietly narrows to what can be measured. Anything that resists measurement is scored as friction. His two-child population policy passed every measure he was watching in the 1970s — it worked, on the numbers he had chosen — and produced a birth rate the country has never recovered. The test had a horizon, and the horizon was not declared.
A generation decision judged on cost per megawatt-hour and construction schedule has exactly that shape. Those are the instrumented quantities. Custody of waste across centuries, the position of a community that hosts a facility for eighty years, and the options foreclosed by committing a grid to one architecture are not instrumented, and will therefore be recorded as working right up until they are not.
There is a larger caution about the lens itself. A criterion that asks only whether something can be made to work, applied by a state that can simply decide, licensed a great deal in his career that was not defensible — including detention without trial for its whole duration, never tested in open court. Energy siting in most countries requires consent rather than capability, and on how to obtain consent he is close to useless.
Answer this next
For the option you favour, name the operating unit and where it is.
If you cannot, you are not comparing evidence with doctrine. You are comparing two forecasts, which is a fair thing to do and a different thing entirely.
