interesting proof of concept, in php, who would have thought :D

Sometimes when doing offensive security work you end up in the strangest environments with limited tools, odd quirks, broken shells, and god knows what else. But you know what is almost always available and just works? PHP.

I wonder if the compiler runs on https://github.com/VKCOM/kphp (a PHP->C++ transpiler)

Are all PHP developers named Rasmus?

you never know what’s going on in someone else’s claude max plan

This is a great example of how people waste our planet's resources faster with AI.

Seriously, in what world do we need a rust compiler in php? I'd rather have cheap RAM and storage, which I can't because of this kind of stupid idea.

I wonder how much energy was wasted on this. How many people got poisoned or killed in mines to create the GPUs that spewed out this useless code.

The fact that we can do something doesn't mean we should. It's time to end this madness.

EDIT: yeah keep your downvotes coming. Ignore the obvious problem. Easy, it's invisible, just don't think about it.

You never know what’s going on in someone else’s Claude Max plan =D

I bet claude was hyping this guy up as he was building it. "Absolutely, a rust compiler written in PHP is a great idea!"

Here I thought my experiment with pushing xz to compress Alpine binaries harder was a waste of time / tokens

https://github.com/mohsen1/fesh

I think people don't really realise that compilers are "difficult" projects in the same way as an appendicectomy is for a skilled surgeon, i.e. the surgery is "routine" only because the surgeons spent decades honing their skills to do these routinely. The hard part was training someone to be able to do that.

Writing a compiler/interpreter is _extremely_ straightforward; a lexer -> parser -> ast -> semantic analysis -> {codegen -> linker | evaluator} pipeline is a very widely understood and tested way to write a compiler in any language, regardless of what language you are trying to compile. The hard part is _learning_ how it works, but after that implementing a compiler is a kind of mechanical activity. That's why LLMs are so great at writing parsers: they can just read the source of any compiler (and they probably read all of them) and apply the same stuff mechanically, with almost a 100% accuracy. We even have formal languages to define parsers and RTL and stuff, that's how "mechanical" the whole process can be.

I'm pretty sure that any skilled compiler dev with the ISO C standard and a few packs of Red Bulls can apecode a working C compiler in a few days, give or take. The hard part isn't doing that, the hard part is the decades of iterative improvements to make it generate extremely performant yet correct code as fast as possible.