The original note, as written at the time
There might come a time when we find adequate ways of shielding against nuclear weapons. However, if that day doesn’t come, all sides of each conflict will continue to seek out fair weapons that each can accept on the battlefield before pulling the last straw.
Ongoing development
There may come a time when we find adequate ways of shielding against nuclear weapons. If that day does not come, every conflict will keep searching for a fair weapon — one that both sides can accept on the battlefield before somebody reaches for the last straw.
That is the part of the problem nobody discusses. We talk about deterrence as though it were a strategy, when it is actually a truce about which weapons are thinkable. Infantry against infantry is acceptable. Tanks against tanks is acceptable. Gas is not. Nuclear is not, unless everything is. The line is not moral. It is mutual.
So the question is what the next mutually acceptable form looks like, and I think it looks like mecha — large, expensive, visible, manned machines.
Not for their firepower. For their accounting. A mecha is so expensive that you cannot field many, so the battlefield is thin. It is so large that it cannot hide, so there is no ambiguity about whether it is present. It is manned, so the decision to commit one is a decision about a named person, not about a munition. And it is slow and clumsy enough that a defender has time to see it coming, which is the single most civilising property a weapon can have.
The conventional objection is that this is fantasy, and that real procurement will not buy it. Probably true. But the argument is not about procurement. It is about the fact that we keep designing weapons with no shared language for restraint, and then discovering the restraint afterwards, in the graveyards. A weapon whose very expense forces both sides into the same small number of decisions might be worth more than a cheap one whose only virtue is that you can afford a great many.