I feel like having an LLM write code in a language you aren't familiar with and then inspecting the results is kind of like hiring someone to speak Spanish for you and then being confused at the weird words they are using. Like, what would make you want to do this?
Not the author but this seems a good approach to me because you learn more about a language from implementing a project in it. This is especially true when you already have experience in a language from the same paradigm (like Go and Rust are).
So getting an LLM to write an example project then dissecting the code and interrogating those choices, seems like a very good way to learn the idioms of another language.
If you believe that then you haven’t spent much time working in different paradigms of programming languages.
Syntax is the easy stuff to learn. It’s any shifts in paradigms (eg pure functional vs imperative vs logic… etc) that takes time to learn.
And I say this as someone who’s written professional software in well over a dozen different languages. So I understand well the challenges learning something new.
I have written production code in about a dozen languages as well believe it or not.
I have also trained people who were good to decent software engineers in other languages to write rust. The syntax is nontrivial for a lot of people. There are a lot of people who gave up trying to learn rust, especially before the rust book became what it is today.
People typically fight the borrowchecker until it clicks. Learning from an LLM and reading only means you have to be as good as the rust compiler without any experience writing the language. It's got to be way harder that way.
> I have written production code in about a dozen languages as well believe it or not.
You said 6 in your other comment. Which is half a dozen.
But I take your point about the syntax being complex. That was the main reason I stopped coding in Rust: not because I couldn’t learn the language but because I didn’t enjoy the complexity. To me it felt like it needed someone to reign in design choices (Python is suffering from this problem now too).
On a slight tangent: one of my pet peeves is feature creep in programming languages. It makes it harder to learn the language. Harder to agree on coding styles in teams. Easier to fuck up and thus requires you to be on your A-game when writing code for it. I don’t always agree with Go’s choices, but I respect that the language is conservative in what gets approved into the language. This is a takeaway more languages need to learn from.
Anyway, back on topic: I don’t agree that the syntax and borrow checker constitutes as “a different paradigm”. But I’ll concede that I might be overstating how easy it is for others to learn these idioms.
Go and Rust have different idioms and syntax. But they occupy broadly similar paradigms.
For example, you don’t need to relearn how to do iteration like you would with a logic or pure functional language. You wouldn’t need to concepts like methods, like you would if you were coming from a stack based language. Etc
I think this comment weasels around the intent of the poster without acknowledging their meaning.
Go and rust have very little in common. If you consider them to be the same paradigm that's fine. But I don't think most people would as rust leans more functional.
“Leaning into functional” isn’t a hard thing to learn. However pure functional is when coming from an imperative language.
And that’s the point I was always making. Rust takes inspiration from different languages than Go. But there is a huge amount of borrowed experience you can lean on when switching between Go and Rust. You’re not starting from scratch.
Perhaps the real problem here is that developers stick to a subset of similar imperative languages and then moan that minor differences are hard to reason about?
Nah, it's an awful way to learn. Especially to learn to be good or great.
When you start reading, it helps to have some guidance towards good and relevant books, from e.g. school, mentors, criticism, etc. Then, when you encounter a "bad" book, you have some benchmarks from which you can build your capacity for analysis and critique. (Testing your analysis and critique with others helps, too.)
If you start with "bad" books, your concept of quality and what's possible is constrained. (Like when teenage boys read Atlas Shrugged.)
Reading slop code is a terrible way to build a mental benchmark for what's good, what's possible, what's elegant, and writing good code that is respectful to your fellow human beings.
I’ve learned a ton of new things this way, even in a stack that I know really well. Ditto on all sorts of new little command line tricks that I was unaware of before.
I mean, it's not that surprising that you'll learn better in a stack you already know well - you know enough there to know what you don't know and need to learn. But if you don't know anything about a language, it will be very difficult for you to sort fact from fiction.
It's one of the best use cases for LLMs IMO. Programmers being empowered to do stuff they wouldn't dare before and/or do what they know, but faster. Having a person who never wrote much code (if any) before is a recipe for disaster because LLMs, even latest models with CC/Codex make mistakes and often code where a happy path (kinda) works, but edge cases don't . You have to check and iterate and specify. But also, programmers (seniors at the very least) have an intuition about how the system should work and they know algorithmization in general. They know how to do a thing in pseudocode on paper. In the end, you become kind of an architect of the system. LLMs give you the ability to choose the right tool for the job even if you have suboptimal or even very little experience with the given tool. There are footguns of course and I wouldn't work on say a system handling client money (banks...) this way, but most uses can take it.
As far as being taught a new language and its ecosystem through an LLM, is SO much faster than reading a book + documentation, it's like asuperpower.
Yeah I have had LLMs write scripts and changes in languages I can't really read for throwaway uses but I have not really found it useful to go and inspect the code because I don't feel I would learn much
It is more like you wanting to build a bed out of wood so you hire a carpenter and watch them and ask questions about every step and maybe help a bit at the end.
After sawing off his arm while assembling an barstool he also gives advice about assembling rocket engines without any loss in confidence. Good old uncle Vinny always there when you need him.
If you already speak French or another Romance language it isn’t a bad idea to just have a conversation in Spanish directly and then ask for clarifications anytime you don’t understand.
Which would be all the time? At which point you might be better served by learning from a source that has any guarantees of being correct and doesn't hallucinate. Like text books that have had several editions and are free on the Internet.
I would be very surprised if you couldn’t figure out what was happening in one C-derivative language when you’re already competent in another C-derivative language.
This isn’t like learning JavaScript and then expecting to be an expert in Prolog.
The first time I looked at rust code that wasn't in tutorial I was pretty confused. Things I thought I understood I really didn't. I knew maybe 6 programming languages including some c. A lot of people struggle to learn rust because it's an ML as in OCAML and really isn't much like C at all.
Some people adapt to it more easily, especially coming from languages like scala but it has a lot of unique characteristics that aren't in C or are even related. Like lifetimes, dynamic dispatch through enums, the borrowchecker, pattern matching, the ? Operator, etc.
Maybe you all are way smarter than me, super possible, but I wouldn't expect much to translate between go and rust. I think some evidence for that is the blog post here...
Yep, I did. While I don't use OpenClaw, I built a small MCP tool for my AI to use Gopher in a minimal harness, and it's been useful. Gopher is almost an ideal protocol for AI, none of the token verbosity of HTML. But I admit in my case, it's mostly being used to access weather data on Floodgap's Groundhog, because the format published on Gopher is much easier to parse & access than the paywalled government APIs in Australia. Claude occasionally uses Veronica to do a search instead of a web search as well.
People really are forgetting how to think. While reading this blog post I almost immediately flipped into teaching confused freshmen taking the course that wasn't their major mode.
A much more charitable reading of the article is "here's some observations I found interesting / different". I met the author at the conf and watched their talk. Their perspective was very much in the camp of exploration, learning, wonder, etc. and not in the camp of criticism of either rust or vibe coding practices.
Interesting that the talk was quite different, because this clearly AI generated article was condescending and infuriating. Quite a difference apparently.
std::expected and the utility functions for it (and_then(), or_else()) are pretty much the same, though? Or am I completely misunderstanding something?
It's true that `std::expected` is like if a C++ programmer saw a type like Result in a shop window and copied the parts they understood from that and so certainly if you're a C++ programmer this is superficially satisfying.
The blog post uses, among other things, the Try operator ? and pattern matching, neither of which are available in C++ and both of which make the Result type much nicer to use than std::expected. There have been similar "I saw it in a shop window" proposals for both these in C++, and I expect that pattern matching in particular will be attempted again targeting C++ 29.
Learning Go after five years of professional struggle with Rust was a relief; Go feels designed for humans to just get the job done. (not a Google fan!) I'll get a ton of downvote for this but it's ok.
I find your experience interesting. People below are saying you can understand rust super easy if you know go. Meanwhile you are saying rust was a struggle and go is a relief.
So which is it?
The answer is rust has very little in common with go. Rust is very explicit and go is not. Some people find the explicit nature of rust and it's guarantees refreshing. Other people find go refreshing because the syntax is more limited and it looks simple on paper.
Yes: I do say Rust was a struggle and go is a relief.
Yes: rust has very little in common with go.
Yes: Rust is very explicit and go is not.
Yes: Other people find go refreshing because the syntax is more limited and it looks simple on paper.
So, you're right.
IMO: Go is a very "productive" and clean lang/platform when comparing to Rust. It's depends what you're using it for. In my case (for concurrent backends) Go came as a bliss. And that was before AI (vibecoding).
The weird-looking Rust isn’t really Rust being weird, it’s the type telling the truth.
Result<Option<Result<Message, WsError>>, Elapsed>
That’s three independent “not the happy path” channels: timeout, stream closed,
and websocket error.
The nicer version is not a cleverer match. It’s choosing a domain error shape
and converting into it one layer at a time:
let timed = tokio::time::timeout(duration, receiver.next()).await;
let next = timed.map_err(|_| ReceiveError::Timeout)?;
let item = next.ok_or(ReceiveError::Closed)?;
let msg = item.map_err(ReceiveError::WebSocket)?;
The ugly line is what happens when you have not decided where to normalize the
shape yet.
It's basically doing the same thing that, say, `return true` might do to indicate a function succeeded, but with more explicit types. However, because it uses `Result`, it can be used with the `try`/question mark operator which can be convenient in some situations.
That said, a couple of the examples here feel a bit strange - they're clever things you can do, but they're not necessarily things you often have to do, particularly for a relatively simple task like this. I think the problem with the author's approach is that they can't distinguish between "weird because Rust is weird" and "weird because the LLM generated bad code", because they (understandably) don't have enough experience in what good Rust code looks like.
it is indeed pretty weird. clippy has a lint against this iirc. it's recommended to just create a custom error type, even if its just an empty struct or a single-variant enum
this lets you implement `std::error::Error`, which you really should to make it less painful when you want to erase the type (`std::error::Error` is `dyn`-compatible)
It's not like people regularly decide this is a good return type. Just because Claude isn't good at designing code or what have you doesn't mean rust is bad/weird.
Sure this is something someone can do but it's suggesting the caller doesn't care about why it failed and doesn't need anything from it's success. It's a choice but it's not a typical one. Maybe the fact that it looks weird and there is no comment is a clue that this isn't high quality code.
People really should be more skeptical of LLM coding. Claude is not as amazing as marketing makes it sound. It is amazing in that it can write code and follow specs sometimes, but a lot of quality gets lost along the way without close supervision by someone who knows better
Sometimes you just want a fancy boolean. The advantage is that Result has all the Result APIs and you can compose it with other Results, but otherwise this is just a success bool.
Probably on topic here - I talk like an LLM sometimes, and parse my points through them sometimes. I’d reasonably use that terminology and think nothing of it as it’s precise and correct. That said, this was partially LLM and my thinking here.