The UX of using Rust is far better than that of C++, but to take C++’s place, it has to become some kind of "AI language".
Why do you think being an "AI language" is an advantage? Mojo tried being exactly that, and I haven't heard anything about it in over a year. The AI market is already covered by other languages (like Python), and many think that the industry is heading for a crash anyways. IMO, leading into AI would destroy Rust, not make it more competitive.
In any case, there are a lot of industries that the Rust ecosystem could conceivably target (assuming the goal of overtaking C++), and AI is possibly one of the worst. Games, low latency trading, high performance simulations, security-critical applications - to me, these are all much more "valuable targets".
(to be clear, I don't see this as a war between Rust and any other language)
9 Likes
I think they meant "a language AI generates well", instead of "a language implements AI well".
4 Likes
What I mean is a more natural language friendly AI programming language. Actually, Rust and C++ are in direct competition.
Yeah. Modern coding has changed from writing code to generating code with writing comments as the core; programming languages themselves have become a middle layer, and comments are the core.
That's just a wishful thinking, at this point — and it only works while “amazing AI programmers” can find willing patsies who may save their slop from unravelling under their weight after being “developed” for a few months.
This being said, we were moving on the road to changing the adoption criteria of “acceptable code” toward more and more sloppy and unreliable code, over the years, training users to accept ever more buggy POS thus the AI-bubble may actually survive for a surprising amount of time, before the whole thing would collapse.
3 Likes
I have to agree, the pace at which AI programming has been advancing is nowhere close to the pace that promises around it have been made. Product after product has been released, with this and that claim, only for it to be "alright" or "a bit better than before". So I really don't see it as the future, regardless of marketing and hype.
But, let's pretend it is the future. Why is Rust a bad candidate? It's one of the few general purpose, performant languages that has memory safety. Its "pit of success" design would theoretically prevent the agents from making certain mistakes. AFAIK, the current mediocreness of AI Rust code (as compared to other languages) is largely just a matter of the training data being much more scarce.
2 Likes
Bad AI programmers do not represent all AI programmers. AI is a great tool, and a good AI model in particular has shown me the promise of natural language programming.
Rust is a great choice—I even think it’s the best one for high-performance use cases, especially vs. C++. But mechanical automation replacing manual work is inevitable in every field. Rust is awesome, and I love it, but I don’t believe programming languages have stopped evolving.
What do you mean by "AI language"?
AI is implemented on GPU's/NPU's. As far as I know C++ is not used to do that. Rather it is CUDA or some such.
Seems a lot of AI training and such is orchestrated with Python. I see no reason Rust cannot do that and have heard there is work going on to do that.
At the user end, we have client code that talks to AI models, which seems to be done in all kind of languages.
Or do you mean something other than all of that?
Currently AI generates code from prompts, that code is in some programming language, including Rust. Seems to me that if the AI was really smart we would not need that intermediate step, the AI would go straight from prompt to executable.
Mojo (Modular) is active and still growing. 25k github stars.
It's great, but extremely limited. And it's not even clear how much would remain immediately after AI bubble would burst (today all the AIs are subsidized heavily, when they would need to provide profits the quality would have to go down and prices would have to go up).
You mean: the same promises that FGCS have shown, isn't it? It was supposed to provide systems programmed in natural languages in 10 years project with massive buildup of parallel computing and corresponding artificial intelligence advances, remember? Sounds familiar, isn't it?
After 10 years of development, approximately by the year 1992 we were supposed to stop using creaky 3GL programming languages (such as BASIC, C, Java, or, yes, even Rust) and go to 5GL languages (which would be very close to natural languages).
I haven't seen even a single AI programmer who doesn't rely on having someone else who would clean up and fix the technical debt after them.
Some of them simply commit enormous amount of workslop and rely on their colleagues. Some create codebases that become increasingly more fragile and less fixable over time in the hope that they would be able to sell their startup before everything would collapse (and then it would be rewritten properly).
None of them are capable of producing code faster than non-AI programmer if they couldn't exploit said non-AI programmers.
It's like “renewable energy” story: it sounds great, in theory, yet even today, 16 years of the start of Energiewende renewable sources still rely on things that are done in a traditional way. None of them may provide energy cheaper than what is done in a traditional way.
Same with AI: it sounds great, some people believe it's the future… yet even 16 years from now it would only be used in a certain niches (prototyping, code reviews, mockups) and would still rely on these creaky 3GL languages (Rust is one of them) and humans who would write code directly in them for the long-term viability.
Well, searching github with created:>2025-01-01 stars:>10 language:Mojo returns 10 repo. Technically it is not totally dead, I guess. (For comparison: C++ gives 4.8k, Rust 5.5k, python 40.9k, Zig ~300, Jai has 8.)
3 Likes
Haha, you've the point. Yes, some rookie programmers might do this. I’m also not a fan of code written by sloppy, imprecise rookie AI programmers. Actually, the arrival of AI benefits experienced programmers the most. They no longer write code from scratch, but instead review code generated by AI, saving a great deal of effort.
It IS great. It's a tech revolution. You're right, some people like me believe it's the future from the bottom of my heart.
It’s not necessary to use prompts anymore. We’re already very close to programming with natural language.
Even if you believe it's the future it doesn't mean it's near future.
FCGS was supposed to enable not just “programming in natural languages” but was also supposed to enable automatic translation that, in turn, would have made it possible for Japanese Businessman to expand into US and India without learning Enlgish…
Well we do have said translation, isn't it? Only… it took us 30 years instead of 10 and was done using entirely different approach.
Same with programming without programming language: we would arrive there… probably in year 2036 or maybe year 2046. And, again, we would use en entirely different approach.
We wouldn't arrive there in the next 5 or 10 years, that's for sure.
At this point we are discussing whether to believe Sam Altman or Demis Hassabis… and given the fact that the latter is the one who prompted the former to declare code red, recently… I kind of trust him more.
I'm pretty interested in Zig, especially its use of compile-time programming.
I get the picture. 
Based on the big tech giants' current investments, I’d say less than 3 years. I’m just afraid we’ll hit the singularity before natural language programming matures, and then AI will take over human society completely.
I’m pretty optimistic about Gemini too. It’s the first time Google’s been ahead in almost three years since ChatGPT blew up.
If mere “investments” were enough then we would have arrived at that stage back in 1992, as per original plan.
Don't worry. Even if Singularity would arrive before the end of this century we wouldn't be there to encounter it.
Do you know why? No? That's because Google always knew that what Sam Altman promised would never work.
They weren't investing money in the scam that is ChatGPT-style LLMs — and were surprised by the fact that pattern matching without any real understanding can do so much.
But, of course, Google always had the resources to duplicate the Sam Alman driven scam… they just weren't trying to create the next investment bubble, they were trying to make AI work. And the pesky thing refused to work.
Of course when another Tulip mania have started they couldn't explain anything to anyone… they started selling tulips, instead. Which was smart of them, but doesn't mean we are anywhere near the breakthrough needed to make AI reliable — and without reliability the dreams of coding large programs without writing code and controlling every line evaporate.
We shouldn't be thinking about Singularity, at this point, but about the way to survive the Greater Depression (as in: something much worse than Great Depression). Greater Depression wouldn't be caused by AI scam, rather the AI scam was seen as the last, desperate, hope to avoid it… that is why trillions were poured on it.
It was never going to work, not without new scientific breakthroughs that may take 10 to 20 years if we are lucky (or 100 to 2000 years if we are not). And when the money river would dry up… it would be interesting to see what would we be left with.
But the last thing we should worry, at this point, is about fate of Rust in the world where programming is done without programming languages.
We may worry about the fate of Rust in thew world where all the real interesting things are hidden behind paywalls and the “free” internet only contains useless AI slop… that's possible future, even if unpleasant one. The arrival of Singularity in next 10 years? Fuhgeddaboudit.
2 Likes
If you are not coding with AI you are like the horse rider who had a skill that became irrelevant in a moment of a momentous Industrial Revolution - people stopped writing letters because of email - people will stop coding because of AI - AI will code - in 2026 we will all orchestrate, supervise and analyse and provide support with the help of AI and never write a single line of code....change is coming...AI slop is the same terminology the horse riders gave the early car editions....let's all accept that the world has changed...c and c++ are dead due to rust....and python can never harness AGI...Rust is the future and you better learn how to vibe code rust now!
1 Like