Lex Fridman: evaluation

Transcript context

…There’s a lot of focus on the self-importance rather than solving real-world problems. It reminds me of the Sinclair C5. Sir Clive Sinclair of Sinclair Computers built a car, and he said, “Oh, everyone will be traveling around in one of these electric cars.” And it was… Rust reminds me of that, where I think the community doesn’t quite understand that in order to get people to move, you have to build something that’s as good as, if not better than what you have now. Yes, people are doing Rust rewrites, but if they only do 85, 90% of the feature set of what we need, like things like coreutils, that last 1% takes 99% of the time. To use Elon’s famous quote, “Prototypes are easy.” Like this kind of stuff is easy. But this, to get a real electric car, you have to make a car as good as, if not better than what we have now, and Rust isn’t in that stage yet. I think we’d– I don’t think anyone would object to seeing Rust code in FFmpeg, but it needs to work as well and support the same unit testing as everything else. It needs to be flawless. It can’t just randomly break. They can’t just randomly break ABI when they want to. It needs to have, I think, more– I think it still has only one compiler implementation. So it’s got to be as good as, if not better, and saying, “Hey, here’s my utopia of memory safety,” isn’t enough, even though we probably all agree that that’s the goal. So I’ve done a ton of Rust, and the two major topics I had was adding Rust modules inside VLC. One of the reasons VLC got popular and which was one of the main architectural decision, is that VLC is a very small core and a ton of modules, right? And so you can write modules in C, in C++, in Objective-C, and anything that is basically interoperable with C. And so we did some Rust modules, and so I have experience on that, and I wrote some of it. And also, like, my new startup called Kyber, is an open source project mainly done in Rust. What Rust is extremely good in, in the sense that it’s a better C++ that cares about memory and allows you to do things about memory ownership that no one else can do so far. However, it’s great when you start a new project from scratch, and you do everything in Rust. But it’s very not good when you interop with existing part. And some part of the Rust community believes that they need to rewrite everything, and everything will be better with Rust. And the answer is like, no. Like, I’m almost always, in all my years of being engineer, manager, CTO of startup and so on, don’t rewrite, right? Is that– That’s the, that’s the initial instinct for a lot of people when they show up to a code base, probably before LLMs, is like… probably because they don’t understand the, the, the wisdom of the way things have been done in the past. They say, “Well, we need to rewrite it.” Hence why there’s a thousand JavaScript frameworks. But the reason is the following, and this is very important to understand. It is an order of magnitude easier to write code than read code. And you see that also with LLM. They can write code, but analyzing is a lot- … more difficult. And so when you arrive to a very complex piece of code, right? You don’t understand it, right? Because it’s so much more effort to understand the code from someone else because you don’t have the thought process. And often I joke about some languages, mostly Perl for example, which has very complex syntax. And imagine I am at my maximum intellectual efficiency in programming, right? And I write the best code ever. I will not be able to understand it myself six months later, right? Because reading code is more difficult. So very often you arrive, you don’t understand all the wisdom, all the business logic, the reasons that were done that is maybe not documented. And you say, “Well, I’m going to write it.” And the thing is, no, you don’t, right? Because that’s, as Kieran said, right? I’m going to rewrite coreutils in Rust. And then, of course, you arrive very quickly at eighty percent then ninety percent, takes a bit more time, and then you got the last ones, right? On the other side, right? So for new projects, it’s great. Everything related to parsing files networking because of the memory checker, boundary checker, it’s amazing, and there is nothing else. To answer a bit differently for us, imagine I take a piece of software like dav1d or x264, right? Which has a ton of runtime in assembly, right? I rewrite the C part in Rust, right? So it’s more secure. Yes. But then you arrive into the assembly, and you can jump anywhere in the memory because we are doing handwritten assembly. So even if I rewrite the C part in Rust, for security reason, you break all the security when you write handwritten assembly because we can jump anywhere. So in my opinion, we need to do something that is secure assembly, right? So which is compile time, check the assembly, which is similar to the checkasm projects that we’re doing on dav1d and x264 with VideoLAN, is to start instrumenting your assembly at compile time to check that it’s not jumping anywhere in the memory. Because else you might rewrite a part of C in Rust, but if you want to have the same performances, you’re going to have inline assembly, and so you destroy your whole security model. So that’s a bit what I think about Rust. No, I just wanna… I would say on a personal level, I’m so in awe about assembly. I actually— Once in a… It never gets old, the speed improvements to show sixty-two x. So there are months, on a personal level, I run our internal test suite at work and just see I’m still in awe at the gains we have.…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.