It's time to stop using slow programming languages. I don't think you understand the magnitude of this situation. Let me explain.

Not quite.

i will still use Go

That's the most made up comparison I've ever seen

Can you do assembly vs C vs Rust vs Zig vs Odin

Please don't use DHH benchmarks as evidence of anything they are incredibly flawed

There's no way Rust is that much faster than Go.

Most of the real application barely compete for the CPU. They are hungry for the database, so bottleneck isn't how fast language is but it depends on external network I/O.

not impressed, each programming language made for different purpose.

You've got big balls and I cannot lie Your other coders can't deny That when a benchmark shows Balls go'n to and fro you get sprung

what are we watching here? an animation of a bad benchmark?

I think it is rather time to stop posting biased results. Here is the benchmark if you allow Elixir to have equally many commits as Rust:

Ok, now add database read to every request, read that cannot be cached and some time consuming algorythm - maybe llm or system one model, to simulate real data processing

Go all the way

the ball man strikes again

Now the question is, are you reading the code Ben?

This oversimplification is common but it’s important to know “context” aka what type of “workload” is being done. For CPU or GPU intensive tasks or embedded real-time tasks, a manual memory management language like C, C++, or Rust works best as your graphic implies. So that extra complexity is worth it since each little performance difference matters a lot and adds up. BUT for Web Servers/Applications where disk and/or network IO is the primary bottleneck and network IO latency is Orders of magnitude more than memory or CPU/GPU compute related latencies, the graphic above is misleading. Using an automatic memory management language like Java, Go, or C# you can have developer productivity and better code readability at the cost of latencies during GC pauses or other things like interpretation in the case of Java (JIT helps with this though). But since the network latency is the main slowness all services encounter, the slower performance compared to C/C++/Rust is almost negligible. Of course with AI I guess everyone will say those readability and simplicity differences don’t matter so just use Rust which is also valid I suppose as long as you know what the code does.

Its time to stop feeding egos of clueless guys. Dont give them any stage.

Yes and a lot depends on the libraries also

balls have dropped

There are some other fast languages missing, don't you think? And they happen to be the most used ones. Nah, just saying...

If you want to compare languages, not framework, then the results are quite different

Thank you, this was the last thing that was missing from this Vibe benchmark discussion. Hopefully people will move on and start building useful products instead of vibe optimizing everything.

ill remember that whenever i need to make 30000 requests per second.

Rust is a NSA Language.😂

God the amount of parroting clueless people in this site

no gatekeep this let them eat ruby

the only guy to consistently take his balls out to explain

It’s time to stop glorifying overengineering for the sake of nanoseconds when 98% of the market actually needs faster development, maintainability, and reasonable costs.

What did Go do to deserve these balls

Its not only about performance, its also about power usage. The question is how I/O bound these specific benchmarks are? If you fetch things over network (not just local) then the time it takes to fetch things makes the CPU time spent just a fraction regardless of language. However given that an AI is the one building the code just pick the most efficient and type safe language which in this world is Rust. Language ergonomics is no longer an issue.

I think your elixir is wrong. My erlang on a 5 year old shithouse machine does more than that.

who cares about slow software who cares about few bugs here and there who cares if you don't understand your vibe code Idk where we're going with this shit?

I don’t think you’ve read the true backstory of how that benchmark creation was a joke.

Why aren’t you thinking more aggressively about the chain of causality? For example, instead of just accepting Rust as it is, why not ask how we can make it unconditionally fast and as secure as possible, and how we can catch and solve potential problems during compilation before the code ever reaches production? Let’s put more intelligence and verification directly into the compiler. The compiler could even actively test for potential attack vectors as part of the build process. We could go deeper into Rust’s libraries and address the real critical bottlenecks at their source. Maybe we could even create a language that is better than Rust. Don’t waste time on the surface. Focus on the real bottlenecks. Go all the way down to Rust’s core, hardware, materials science, and physics. That’s where the real breakthroughs are.