Yeah they do have a similar mechanism after the new typesystem changes, but not the exact same thing

Network effect. Everyone uses TS because everyone else does. JS is such a terribly designed language that no amount of tooling or bolted on typesystem can salvage it. TS is an impressive piece of tech nonetheless. WASM need to happen so that I can write Go for frontend.

Pessoal, acho que decidi falar sério neste post: A área de desenvolvimento de software, principalmente no Brasil, já passou por vários ciclos de "os devs vão acabar". Desde 1970 se tenta promulgar e impulsionar geradores de código, no-code, etc... De todas as interações anteriores, essa é a mais preocupante: Antigamente era melhor ter MENOS código. E o custo disso era trivial. Agora você precisa gerar MAIS CÓDIGO (pq a IA simplesmente faz isso). É extremamente contraprodutivo. Sim, a ferramenta é ótima. Mas é um passo contra todos os avanços dos últimos anos. Porra, o DHH já tá usando TOKENMAXXING como argumento pra acabar com typesystem

java never has had a good dynamic metaprogramming system: reflection works but lives on top of a static type system, which is why codegen is so prevelant there are tools that plug into the compiler & classloaders though manifold.systems plugs into the compiler and implements the open typesystem concept from the guidewire programming language gosu (gosu-lang.github.io/) which is the language I learned from scott taught me more about PLs in a few months than two years at stanford taught me

In rust you need to explicitly warp everything, and wrapping makes the types incompatible with the unwrapped types, because Rust's typesystem is a castrated version of ML's typesystem from 1973.

To be fair, Rust is the only language they could have "rewritten" it to such that the argument still kinda stands. Rust is the only language mature enough, that gives you the same level of control (although with a different default, similar metaprogramming capability (although differently executed), same-ish control over allocators, and arguably better typesystem. Now, oneshotting an AI rewrite and just merging fills me with dread and feels like a thinly veiled Anthropic publicity stunt, but bidirectional rewrites between Rust and Zig are one of the more plausible ones. I recall Roc lang did a similar thing, but in the other direction (Rust->Zig) and without AI, which is probably a better practical case study

Not sure what you mean by "follow", but certainly, your puny typesystem will make your life difficult. Without recursively constrained types, you'll have to resort to encoding dynamic data in a rigid static typesystem.

Can confirm this. I can't count how many bugs related to cancellation, goroutine leaks, incorrect graceful shutdown sequences (causing lost msg or requests) I’ve seen and fixed (I’ve worked at companies using Go for 8 years now). Concurrency and the typesystem is terrible.

c# is perfect for ai development - the typesystem and the large pool of built in features keeps the ai on track without boiler plate dependency work

Adotar uma linguagem com tipos dependentes vai permitir a gente a eliminar bugs em vez de corrigir eles. Se o typesystem de Rust é um beneficio pq não colher mais beneficios ainda? A gente pode usar o LLM pra escrever provas em vez de testes

which typesystem do you prefer, ocaml or ts?

I had opposite experience: in the statically typed language, I trusted the typesystem too much so didn't test enough. Porting to dynamically typed, I had to test a whole lot and found bugs in the original, while fixing the static type issues was a few tests away.

99% of it is hype, and most of the benefit comes from Erlang OTP, not BEAM You really really want some form of verification and thus a good typesystem for LLMs, so Elixir doesn't really cut it & I doubt it'll scale. But Gleam OTP may be really nice if it exists (not sure if OTP works with Gleam yet, haven't been following for a bit)

I disagree: Types help productivity if you consider debugging and reduction of boilerplate/code repetition to be productivity. (Which is not to say that having a strong typesystem is the only way to reduce boilerplate)

Bring it. The TypeScript gang stands in a bizarre spot in the types debate, since their typesystem is intentionally unsound I sort of like that philosophy btw. You get something from the types, but you can tell the types to go fuck themselves whenever you feel like it

More precisely, it failed because it is statically typed, and no two languages (including XML) can agree on a non-trivial static typesystem, whereas JSON is all about agreeing on a trivial static typesystem, aka dynamic typing.

I know, this is extremely bad design and it's only the grammar not the typesystem. Variable a still has type "pointer to char" and there's nothing special about the pointer in that type construct. And this is one reason good style guides forbid multi-var declarations.

He then asks for a demonstration in typescript, like typescript is the only typesystem

Not doing much pure FP these days but still miss the power of such typesystem algebraic compositionality a lot .. The Applicative instance says “combine independent checks and accumulate errors (Semigroup e)”. makeFullTransaction <$> ... <*> ... works smoothly because constructors/functions are naturally curried.

Le langage a énormément évolué, toute la partie "bloatware" (gros serveurs d'application type jboss, conf xml etc) a disparue, le langage est bcp moins verbeux, on a un peu de prog fonctionnelle (stream avec mal/reduce etc), un meilleur typesystem, des data class (record) etc