This is a persuasive article, but it is still not clear whether Rust will achieve a self-sustaining mass of developers. That is, precisely: enough that each new person who masters Rust will be able to find work writing Rust. (To be clear, it would be good if this were to happen.)<p>The deepest observation in the article is the remark about the cognitive load imposed by need to avoid pitfalls inherent in the programming model.<p>I coded K&R C in the 80s, and while I never shipped an argument-type bug, avoiding them was a continual distraction. I coded C90 from 2006 to 2012, with a similar cognitive load I was happy to leave behind, although again I never shipped a bug caused by its failings. (We had bugs, but they were specification bugs: building the wrong thing.)<p>Writing C++ post-14 has been an increasing pleasure. The cognitive load noted is present at times, but decreasingly so as the Standard Library and other libraries get more powerful.<p>I am disappointed that the author found it too hard to implement a correct program with C++, but more disappointed that he chose to blame his tools for his failure.<p>It has become fashionable to insist that correct programs cannot be written in this or that unfashionable language, but in fact a great many correct programs have been written in all of them, not excepting octal machine code.<p>What varies is how much work is needed, and how many people can do it: it is certain that, for any chosen language past or future, most people cannot or will not write a correct program in it. Rust is not different in this.<p>Many people coding C++, and perhaps most coding C, are not writing correct code, and have over the years written a very great deal of bad code. Will those same people succeed in learning Rust, and learning to write good code in it? It appears that the people currently adopting Rust are more skilled than is typical, so it seems hard to generalize their results to more typical programmers.<p>C++ and Rust embody different choices in emphasis: Rust, safety; C++, library power. Stewards of C++ have decided the best way to get substantial correct programs is to enable encapsulating semantics in well-tested, well-optimized libraries that can be used without compromising performance or system architecture. Rust has chosen to emphasize making wrong low-level code hard to express, although its ability to express libraries will only increase.<p>It will not be clear for a long time which will have better results. Will enough people learn Rust to write the correct low-level code needed? Will enough C++ programmers learn to use libraries that make (re-)writing dangerous low-level code unnecessary?<p>What is clear is that the number of C++ programmers is growing faster than ever, and interest in coding to high level C++ libraries, as measured by attendance at an ever increasing number of conferences, is exploding. Even attendance at each ISO Standard meeting is quite a lot larger than at each previous one, now reliably in hundreds. The number of new C++ programmers in any unit time is still larger than the number of new Rust programmers, by a large factor.<p>So, the experiment is interesting and valuable. What we can be confident about, either way, is that, for every C coder, or C++ coder writing low-level, often bad code, switching to Rust or to C++ using reliable libraries will result in more good code. It matters much less which they choose.