Why is it that virtually all new languages in the last 25 years or so have a C-like syntax?

- Java wanted to sort-of knock off C++.
- C# wanted to be Java but on Microsoft's proprietary stack instead of SUN's (now Oracle's).
- Several other languages such as Vala, Scala, Swift, etc. do only careful evolution, seemingly so as to not alienate the devs used to previous C-like languages.
- Not to speak of everyone's favourite enemy, JavaScript…
- Then there is ReasonML which is basically an alternate, more C-like, syntax for OCaml, and is then compiled to JavaScript.

Now we're slowly arriving at the meat of this rant: back when I started university, the first semester programming lecture used Scheme, and provided a fine introduction to (functional) programming. Scheme, like other variants of Lisp, is a fine language, very flexible, code is data, data is code, but you get somewhat lost in a sea of parentheses, probably worse than the C-like languages' salad of curly braces. But it was a refreshing change from the likes of C, C++, and Java in terms of approach.

But the real enlightenment came when I read through Okasaki's paper on purely functional data structures. The author uses Standard ML in the paper, and after the initial shock (because it's different than most everything else I had seen), and getting used to the notation, I loved the crisp clarity it brings with almost no ceremony at all!

After looking around a bit, I found that nobody seems to use SML anymore, but there are viable alternatives, depending on your taste:

- Pragmatic programmers can use OCaml, which has immutability by default, and tries to guide the programmer to a functional programming mindset, but can accommodate imperative constructs easily when necessary.
- F# was born as OCaml on .NET but has now evolved into its own great thing with many upsides and very few downsides; I recommend every C# developer should give it a try.
- Somewhat more extreme is Haskell, with its ideology of pure functions and lazy evaluation that makes introducing side effects, I/O, and other imperative constructs rather a pain in the arse, and not quite my piece of cake, but learning it can still help you be a better programmer in whatever language you use on a day-to-day basis.

Anyway, the point is that after working with several of these languages developed out of the original Meta Language, it baffles me how anyone can be happy being a curly-braces-language developer without craving something more succinct and to-the-point. Especially when it comes to JavaScript: all the above mentioned ML-like languages can be compiled to JavaScript, so developing directly in JavaScript should hardly be a necessity.

Obviously these curly-braces languages will still be needed for a long time coming, legacy systems and all—just look at COBOL—, but my point stands.

teach meta language concepts: what is an operator, literal, constant, statment, control flow. the recursive nature of staments. then go into objects/methods vs structs/procedures. then teach some java. then go into reflection concepts. then use reflection for something simple. then teach a bit of perl. then let them build something in python. Anyone who can pass through that will know how to Program in whatever you give him/her.
I wish my teachers talked about the meta programing, instead on focusing on the minutia.

Since day 1, I've been eager in learning everything dev-related in English, and as it, I've focused on thinking, writing, shitting and eating english for a basis.

Now, where I work, I say, our country native language it's Spanish, but I feel bad whenever they tell me I must write everything in Spanish (code included). Maybe I'm overreacting or something.

So, here's my question for you, devs, when you're working on a coding-related project, should it be on English or not?

How hard can it be to sort content stored in a relational database by a custom meta parameter and restrict the results to a certain language using a very popular content management system in 2023?

After wasting several hours trying to get my head around reference documents, 20 years of anecdotal StackExchange + WordPress.org discussion and ACF + Polylang support, and trying to debug my code, I will now either write my own SQL query or put the meta query results in a hashed object to sort it using my own PHP code.

What time is it now? 2003?

How meta will it be when I get close to finishing the interpreter for the language I'm building, and I use my OWN language for all my future projects?

They keep training bigger language models (GPT et al). All the resear4chers appear to be doing this as a first step, and then running self-learning. The way they do this is train a smaller network, using the bigger network as a teacher. Another way of doing this is dropping some parameters and nodes and testing the performance of the network to see if the smaller version performs roughly the same, on the theory that there are some initialization and configurations that start out, just by happenstance, to be efficient (like finding a "winning lottery ticket").

My question is why aren't they running these two procedures *during* training and validation?

If [x] is a good initialization or larger network and [y] is a smaller network, then
after each training and validation, we run it against a potential [y]. If the result is acceptable and [y] is a good substitute, y becomes x, and we repeat the entire procedure.

The idea is not to look to optimize mere training and validation loss, but to bootstrap a sort of meta-loss that exists across the whole span of training, amortizing the loss function.

Anyone seen this in the wild yet?

I wonder if anyone has considered building a large language model, trained on consuming and generating token sequences that are themselves the actual weights or matrix values of other large language models?

Run Lora to tune it to find and generate plausible subgraphs for specific tasks (an optimal search for weights that are most likely to be initialized by chance to ideal values, i.e. the winning lottery ticket hypothesis).

The entire thing could even be used to prune existing LLM weights, in a generative-adversarial model.

Shit, theres enough embedding and weight data to train a Meta-LLM from scratch at this point.
The sum total of trillions of parameter in models floating around the internet to be used as training data.

If the models and weights are designed to predict the next token, there shouldn't be anything to prevent another model trained on this sort of distribution, from generating new plausible models.

You could even do task-prompt-to-model-task embeddings by training on the weights for task specific models, do vector searches to mix models, etc, and generate *new* models,
not new new text, not new imagery, but new *models*.

It'd be a model for training/inferring/optimizing/generating other models.

I think I've asked this before. Just cropped up again cuz I'm pushed to do some stuff in nextjs

I Wonder how much longer before js framework devs realise they've been reverse engineering the browser this whole time, that the current browser spec was outdated since the dawn of Web fidelity and real time applications

I wonder whether there are some guys who have seen this and are already cooking in the background. The browser still treats the Internet like front end and back end, whereas with the way apps are going (eg deprecation of the front/back end roles), it seems apparent the browser needs to scale up by fading whatever js is now

I'm seeing "use server", which was one of php's infamous atrocities back in the day (lack of separation of concerns, everything in index.php). It's shocking how those who ridicule that language let this fly, but that's probably a separate thread. Point is, a bunch of these stuff done by front end frameworks seem like boilerplate but the syntax is far different from what I remember javaScript to be. I only vaguely recollect and understand what I'm reading

Why not merge all the cryptic syntaxes struggling to achieve bare minimal expectations, into advanced markup language controlled by dom attributes? Overhaul and Rethink client - server communication to fit modern standard. Someone needs to step out of the box and take a good look at the rat race. I find our lives would be made much simpler if api integration into client side behaviour wasn't a separate thing altogether

You have all these funny hoops and precarious bridges to cross. The reality is what we're fighting to overcome is the manner the architecture is setup. We need a Google/meta/amazon/apple to step in with a new browser since it's not a weekend gig and might need their reach to catch on with mainstream users. Sadly, they're the same guys rolling out new js frameworks

The lack of a meta-language in c# can be a pain in the ass, I have to jump through hoops to generate something like python's decorators, not to mention having to generate il to overcome some limitations of reflection when dealing with value types.

Holy crap, Meta Developer Connect keynote. Amazing innovation. This is what Apple **used to be**

Granted, the hardware is not as elegant as Apple but the cost is 1/10 and the capabilities are close, same, or exceed (Llama) what Apple is offering.

Now here is the gut punch, they figured out that the mobile app build system needs to build AR/VR/MR apps. That was Apple's edge.

As a developer, I am not enamored with Swift and it is pretty clear that if I have to change and use a niche language like Swift or change and do dev on Android, to target new Meta hardware and AI... well... lets just say I think Swift is crap from a language standpoint and I suspect it is the reason Apple's hardware uses so much more memory, battery and storage than it should. At the same time Meta's Orion runs on a god damn battery in the early piece of glasses. My AVP's have a huge brick.

#define kApple kGigaBloat

If I were Apple I would be shitting my pants watching this Meta presentation.