Programmers would create a new language when there is a fundamental change in hardware architecture such that the assumptions underlying the old programming languages no longer apply. Java is probably a poor choice for writing software in which most computation happens on GPUs. But I agree that someone (or something) will still be using Java to write new line-of-business applications in 50 years.

Mostly a tongue in cheek reply, but you might be interested in TornadoVM.

(Nonetheless, I agree with you)

> Why would anyone create a new language now? The existing ones are "good enough", and without a body of examples for LLMs to train on, a new language has little chance getting traction.

Compiler writing can be an art form and not all art is for mass consumption.

> Java has won (alongside many other winners of course), now the AI drawbridge is being raised to stop new entrants and my pick is that Java will still be here in 50 years time, it's just no humans will be creating it.

This makes no sense to me. If AI possesses intelligence then it should have no problem learning how to use a new language. If it doesn't possess intelligence, we shouldn't be outsourcing all of our programming to it.

> This makes no sense to me. If AI possesses intelligence then it should have no problem learning how to use a new language. If it doesn't possess intelligence, we shouldn't be outsourcing all of our programming to it.

Perfection. You have made such an excellent. However, I don't want to detract from that but it's like, in reality, this is a completely obvious point but because of this AI/LLM brain-rot that has taken over the software programmer community, writ large, this is particularly insightful. It's also just a sad and unimaginative state we are in to think that no more programming languages will ever be needed other than what currently exists in March 2026 because of LLMs.

> this is a completely obvious point but because of this AI/LLM brain-rot that has taken over the software programmer community

This is really insulting. I think you're wrong, AI agent programming is very good now, and you will have to admit it at some point.

Insulting to whom? Are you anthropomorphizing LLMs. Also, you think it is insulting it is a bad idea to complete outsource software development to it. If that’s the case then being insulting sounds like a good thing.

>Insulting to whom? Are you anthropomorphizing LLMs.

Obviously not. Insulting to programmers who use them, hence I object to the term "brain-rot". Clearly the insinuation is we're just suffering from brain rot.

Intelligence is not a well understood concept. "AI" is also not a well understood concept - we have LLMs that can pick up some novel patterns on "first sight", but then that pattern takes up space in the context window and this kind of learning is quite limited.

Training LLMs on the other hand requires a large amount of training data.

Why would anyone play chess in 2010? The drawbridge is rapidly being raised on human competitiveness.

The vast majority of programming languages ever created never aspired to win and I don't think that's going to change now.

> Why would anyone create a new language now?

I'm writing my own programming language right now... which is for an intensely narrow use case, I'm building a testbed for comparing floating-point implementations without messy language semantics getting in the way.

There's lots of reasons to write your own programming language, especially since if you don't care about it actually displacing existing languages.

Most of my 20 years of experience is Java. Now 3 years into a new job mostly using Python to build microservices. I feel much more productive using Python (plus uv, ruff and mypy for fast repeatable package management, linting and type checking). I see Python having a trajectory to keep improving and gaining more adoption - eg Python keeps growing in popularity https://survey.stackoverflow.co/2025/technology/. It will have real threads soon and better type checking with ty (thanks to astral who also make uv and ruff). Python gives an incredibly tight integration loop since you don’t wait on compiling. Our kotlin projects are always at least 2x slower for ci/cd builds. I like typescript, but working with it in IntelliJ is incredibly slow.

Java compiles as fast as go, so not really sure what's the problem - you can basically use it in an "integration loop" with no problem.

Python's great until you have to refactor a large code base

> Why would anyone create a new language now?

Same reason you'd ever create a new language — to start anew, breaking free from the shackles of backwards compatibility of the old language, having learned the lessons of the past.

The AI angle makes even less sense — surely we will want to create languages that are tailored for them.

What advantage do old languages have that can’t be overcome or at least reduced to insignificance?

The 50-year head start in training data, runtime, and ecosystem? That may not be much, because LLMs are rapidly accelerating software development. LLMs can also generalize: take what they learned for one language and apply it to a “similar” language (and I think most modern languages are similar enough for broad effective translation: all have records/unions, objects, functions, types, control-flow, exceptions, and more). Some fairly recent languages (e.g. Rust, Swift) already have comparable runtimes and ecosystems to older ones, from human-driven development acceleration and concept generalization.

In a recent study, LLMs solved the most exercises in Elixir (https://autocodebench.github.io Table 4). Anecdotally, I’ve heard others say that LLMs code best in Rust, and (for UI) Swift. More importantly, I don’t see an older language advantage that is widening from LLM use; an older language probably is better for most use cases today, but any written code can be translated or regenerated into a newer one.

> I don’t see an older language advantage that is widening from LLM use

A classic one is C++. Microcontrollers like esp32 cost about $10 these days, for a machine more capable than an early PC.

One downside though is that you typically need C++ to program them, and the barrier to entry with C++ is very high, especially for non-programmers.

LLMs remove that barrier so that anyone can create powerful embedded devices - programmed in C++ - without knowing anything about C++.