Man I feel old, I remember the days when one wouldn't be caught dead having not anything at all against JS!

I imagine archaeologists of the future scratching their heads saying something like:

"What were those primitives thinking, not yet having anything for JS, the Double Helix Code of Zombotron The Magnificent! Savages! All Hail NPM!"

I "C" what you did there!

Lemon, man! Watch what you twype!

No question, given the choice it will have to be a skilled worker with AI. If the company wants to survive that is. Remember the saying "a little knowledge is a dangerous thing".

Oh, crud!

That's the problem, when a company is betting its existence on a single product offering or customer. When something changes, their existence has to change too.

Diversification is not just for portfolio managers, it also helps companies survive to do business another day..

We're talking qualitatively, but at completely different scales. The comment chain I replied to posits an equivalence between trusting a complex instruction given as an English language prompt to an LLM code generator, versus trusting a chain of well documented API calls with deliberately restricted semantics. It's completely different.

Even when you add in the possibility of deception, there is no equivalence of trust. Vibe coding intrinsically doesn't follow the expected engineering patterns and practices which are used for large projects. Yet these patterns facilitate the investigation, and discovery, of deception. Without them, the task of verifying a system is orders of magnitude more difficult.

Thanks for the (two) replies. Yes, I was confusing bytecode and machine code, or rather, I was thinking of two potentially different machine codes with isomorphic instruction sets, but potentially different resource availability (RAM, number of registers).

Speak for yourself mate. From where I'm sitting, the Chinese look pretty sane compared to you. But keep making shrill noises and telling us how everyone you don't like is evil, I'm sure it will just turn everyone against you.

If jwz were reading this, he'd agree with you *righteously*!

What if you translate from one bytecode to another? eg into MMIX bytecode.

People didn't like embedded FLASH first.

You have to remember that data science majors are not programmers. They don't care about procedural correctness and are much too green to understand how their data gets into a CSV. They just want to apply some recipes they were taught in their last year of school.

This means that python/basic IS the right tool for the job, if the majority of the worker population has those characteristics.

In the case of R, it's because it's the only place where high quality research level statistical algorithms are built en masse (ML libraries in python are not a substitute, they tend to be built by non-subject matter experts who don't even know that there are corner cases)

You can trust your OS and hardware because it wasn't vibe coded.. Try again.