Ah yeah Yuri Vishnevsky's criticism of Julia. That was a bit topic. I wrote a response to that:

https://erik-engheim.medium.com/why-i-still-recommend-the-julia-programming-language-d2b84a747a5c

Once example from my response which I want to highligh is the Apache Arrows project. It has been implemented in multiple langauges. The C++ team began first and create the reference implementation.

Yet Julia with the least amount of resources and just one guy beat everybody else making the first full implementation of Apache Arrows.

Part of the reason for that is the extreme reusability and generality of Julia code which allows developers to leverage a ton of existing code. The Julia implementation is a tiny fraction of the C++ implementation. It is so small you may think it is a joke.

You see the same with Machine Learning libraries such as Flux. It is so small that many people didn't bother to try using it because they didn't think it was done.

So what does any of these examples have to do with correctness? It is an illustration of why the correctness assesments are doing a Apples to Spaceships comparison. Very generic code is harder to test will work in every combination.

If you removed all that Julia flexiblity then sure, things would be easier to test and you might have more robust code.... in theory. Except you would have been left with a tiny fraction of the functionality Julia is now offering. Given how little resources the Julia community has, it is absolutely astonishing how much functionality they have created.

There are examples like this from TensorFlow I belive as well where a couple of Julia guys on their own spare time outcompeted big paid teams at Google doing similar stuff with Swift.

Julia has rough edges today I believe because it it punching way above its weight class. It is a very young language and a relatively small community which still has pulled of creating a massive amount of functionality.

I think Yuri Vishnevsky criticism is relevant and must be taken serious. But his conclusion that somehow Julia is fundamentally flawed and unfixable is just way off the mark IMHO. His assement does not take into account producitivty.

I will concede this though: Julia probably does suffer some from being so heavily in Academia rather than industry. But this could change as the language gains broader industry acceptance.