Accusing someone of going too fast is the most suspect argument there is. It has been used to hold back every innovation, and it is almost always invoked by those whom speed inconveniences. Yet when leading mathematicians use it while explicitly acknowledging that the models solve open problems, we have to look for what exactly they are pointing to.
This morning we reported their statement. Let us take the time to work through the argument.
What slowness used to produce
Let us distinguish three things the word covers, because they are not of equal value.
Slowness of production. The time it took to find something. This has no virtue: nobody mourns the years spent searching for what can now be obtained in weeks.
Slowness of formulation. The time between the moment something is found and the moment it is published. This is where most of the invisible work happens: isolating what is new, naming it, showing where it comes from, making it reusable. A raw result gives an answer; a worked result gives a method.
Slowness of absorption. The time a community needs to read, verify, challenge and integrate. We described this in our article on the acceptance of proofs: two years of formalities, and that is not bureaucracy.
The criticism applies only to the last two. And it applies rightly.
Why speed destroys the second
Here is the mechanism, and it is economic before it is intellectual.
The work of formulation has no value for the person who publishes quickly. It does not change the result, it does not make the headline, it delays the announcement. In a competition where priority decides, it is rational to skip it.
But that work has value for everyone else. It is what turns an isolated result into a tool others can use. A discipline that accumulates unworked results accumulates facts without method — that is, it does not progress, it lengthens.
This is exactly the mechanism we described in our article on automation: a task whose usefulness benefits others rather than the person doing it will be eliminated first, because nothing in its incentive system signals its cost.
There is a difference between going fast and producing more than can be absorbed. A discipline has a finite number of people capable of verifying a difficult result — sometimes only a few dozen worldwide. If the rate of production exceeds that capacity, a growing share of what is published will be neither verified nor contradicted. These results pile up in a grey area, and further work is built on top of them. This is what we noted in our article on proof: a result is only established if someone has been able to challenge it.
The objection, and what it is worth
Slowness can also be a privilege and a protection. A community that controls the pace controls who gets in, what is published, and which approaches are legitimate. Any defence of the long haul must be examined against that standard.
There is, however, a simple test for distinguishing useful slowness from defensive slowness: does it produce something verifiable? A delay used to reread, formalise and isolate methods produces a better object. A delay used to wait one's turn produces nothing.
By that criterion, the mathematicians' argument holds, because they point precisely to what time produces: the writing up, the isolation of ideas, the citation of prior work. These are things one can observe in a paper, or observe to be absent.
What could reconcile the two
There is a lead, and we came across it this week.
Formal verification removes precisely the part of slowness that serves no purpose: the line-by-line rereading in search of a logical gap. A machine does it in minutes.
What it does not remove — and must not remove — is the work of interpretation: what this result means, what it contributes, where it comes from, what it is used for next. That work is human, and it remains slow because it is difficult.
The right division of labour would therefore be: accelerate verification, protect interpretation. That is what no current incentive encourages.
What to take away
Slowness is not a virtue, nor is it simply a cost to be eliminated. It is the name given to the time during which a result becomes comprehensible, verifiable and transmissible.
What is at stake does not, moreover, concern mathematics alone. In every field we produce more than we can read, verify and integrate. The question is not how to slow production down, which will not happen. It is what to do with what we have no time to understand — and for now, the answer is that we stack it up.