Ok, nuance - it obviously matters, but in what way does it matter? 10 years ago teams would debate about languages and would almost never consider a huge migration because of high costly it’d be. Now, has this changed? Agents know any language. What would make someone switch?

😋

Oh

Language doesn't matter until you talk to someone working on the BEAM There's drawbacks, but the runtime and declarative approach make it fantastic to build with Learning curve is steep, but agents are very good at writing it now

Only matters for a couple reasons:

A TC bump

I'd say any language that compiles and is relatively memory safe (and not esotheric) Two reasons: 1/ shifting left, you catch a lot of issues way early on during compilations. 2/ There is absolutely no reason not to! In the past, more approachable languages were chosen due to simplicity. Now, well agents write it for us, so yeah you get the benefit of a great language. Why not to choose these languages over something very mainstream? Lack of needed libraries (ie Rust sucks at front end stuff). If you are really not familiar with the language, it may cause you issues in trusting the agent output.

I choose to optimize for happiness. Ruby

if you pick Haskell as PL, the quality of unsupervised generated code will drop tremendously. It is being improving but still too slow. Cost of the supervision is unbearable.

Do I need to build and deploy faster (eg, Rust or TS builds are too slow)? Move to Go. Does it need code to metal performance, memory management and build times don’t matter? Move to Rust. Do I need to integrate with Microsoft specific systems? Move to C#. Build for the use case instead of what you’re most comfortable with, that being said, if you don’t understand the code you shouldn’t deploy it. If something goes wrong and you don’t understand it, you’re in trouble.

If I’m using anything other than typescript.

matters more than ever. choosing a programming language never was (and never should be) just about personal preference, but about what the job needs: performance, readability, security, types. it’s like choosing the right infrastructure for a specific workload.

Tools matter. Programming languages, like any system, are shaped by different design principles. Even when they serve the same purpose, the way they achieve it varies. And that difference matters.

The language chosen affects the choices the AI tools make. You should pick a language with depth in your problem space unless you want to change something specific. For example I wanted to bring more visualization features to data engineering so I use JS not Python.

It does but I would just stay with Erlang because it's unique features and unique runtime give a advantage for agents too

Yes it does! Programmings languages are purpose oriented. You technically can write anything in any language but, you don't write a kernel in JavaScript and a webapp in C

It matters, but only in the ways unrelated to authoring. Go has dead simple deployment. If you prefer the performance characteristics of a JVM, you have a few choices. Rust (from what I understand) is (most) memory “safe”. Python is still king for ML related workloads because of the ecosystem

Ya it does. A lot depends on the purpose of work and capabilities of the tool/lang

I’d switch to a language that offers a substantially higher level of abstraction, with clear and verifiable semantics, while still compiling predictably onto existing infrastructure. Even if agents write the implementation, and even if I rarely read it directly, I still need to express what the system should do and verify that the result preserves that intent. The language should shape the vocabulary I use to write specifications, making as much of their meaning, constraints, and guarantees machine-checkable as possible. Language choice still matters, but increasingly as the interface between human intent and generated systems rather than as a tool for writing code by hand. By “higher level,” I mean something like SQL, but for the entire system, not just queries. That language doesn’t exist… not yet, at least :)

Nothing - only performance- customer need - developer experience (translating to velocity or stability)

I don’t think it matters anymore but I still choose the ones I’m comfortable with. Makes it easier for me when I need to dive into something myself.

IMHO a lang with a good compiler make the diference, if someting goes wrong the quality of errors decriptions make the diference

Yeah it matters for the iteration speed of your agents. Frontend: Typescript Backend: Go

baml!

language choice matters not because agents or humans write it, but because of benefits it provides for your codebase/product and if migration costs covered by profits. idk, but it was always like this, no?