s/frequently/likely

(Large language models are not plain Markov chains, contrary to popular belief)

Having parameters for "temperature" and so forth, while it might provide sometimes useful functionality, does not change the basic nature.

To be fair this is a problem of the Rust web ecosystem. Include a few de-facto standard Rust web crates and you're up to 200 transitive dependencies

This story brought me to tears within two paragraphs, wow

I thought it's to transmit good reputation to Meta

E.g. when people trust WhatsApp, and then see it's from Meta, they may trust Meta more too

X is the client and server for rendering pixels on to a screen, which is mapped to one of more monitors. It provides the most basic, bare bones way to draw a window to the screen, and that’s about it.

It’s not even a GNU/Linux thing, X predates Linux, it’s more of the standard system to draw a window on Unix-like systems.

X11 apps are GUI apps uncontrolled by/unaware of a desktop environment, theme engine, or window manager. They appear in a minimalist, maybe even brutalist, aesthetic. You can launch a single X11 app from the console, and the resulting window will be your entire graphical display/instance, returning you to the console after the program terminates execution.

Maybe another way to think about your question might be, think of X vs Wayland.

Interesting

> You can launch a single X11 app from the console, and the resulting window will be your entire graphical display/instance, returning you to the console after the program terminates execution.

Is this not possible with GUI apps using e.g. Qt or Electron?

You interpret X11 apps as a technical descriptor, but I think it's more a description of the ecosystem in which these apps evolved and whose culture they represent.

I see, thanks for explaining

Also most (all?) of these apps use xlib directly, rather than going through a toolkit like Gtk or Qt.

There are a lot of toolkits besides GTK and QT: XT, Motif, Xaw(3d), xforms, xview, wxwidgets.

A lot of these are built against Xlib directly, or X11-only toolkits such as Motif. Porting them to other window systems would require significant rewrites.

Exactly, and the reason 293k boots were needed anyways is because the bug was hard to reproduce and needed thousands of boots to potentially trigger

That’s why I always push telemetry onto a team. Concurrency and resource bugs are extremely expensive per iteration. Anything that can hint that the problem you are certain happened somewhere in a two month period was more likely to have happened in a particular week saves you a ton of work.

The human brain struggles when it has to execute almost, but not quite, the same exercises over and over again. It becomes a blur and after a while you can’t recall if you ran all of the steps this cycle (if you can’t automate the test well) and all it takes is typing “bad” when you meant “good” to end in failure. Which doesn’t sound like a big likelihood but when you’re testing a negative it’s much easier to get your wires crossed.

As per Rust reference, even a capture-less closure is a closure and distinct from an anonymous functions.

Also, your arguments only partly apply in Rust. Rust doesn't heap-allocate closures. And you also often don't have to deal with lifetimes - a closure that captures variables by move or copy is perfectly self-contained

The difference between a closure that captures and a closure that doesn't is like the difference between `(T)` and `()` - same kind of thing, so it adheres to the same terms and behaviors

Other comments are saying non-capturing closures are not closures. If we want to have this pedantic argument of definitions: proto_lambda is actually right, with Rust's definition of closures.

According to the Rust reference:

> Closure types > > A closure expression produces a closure value with a unique, anonymous type that cannot be written out.

https://doc.rust-lang.org/reference/types/closure.html

Even a non-capturing closure is written using a closure expression and produces a unique, anonymous type.

People keep trying to use those LLMs as if it were a search engine but it's not. The value in ChatGPT is its faithful recreation of human common sense.

Use it to hold a conversation, to ask feedback on a text you wrote, to come up with ideas for something. Don't use it as Google and be dismissive when it's not the universal information retrieval tool it's not meant to be

Right. It's almost as if your average human has little creativity, indeed less than what ChatGPT can demonstrate, even though that magical "creativity" is supposed to be the current Big Thing that separates us from the machines. (In a year, once everybody is forced to agree that LLMs exhibit real creativity, not just regurgitation of their training data, the next Big Thing is going to be something else.)

Yeah it's really the back-and-forth nature of it that I find to be a big improvement over searching-and-clicking.

Sooner or later most of the people will use it as a better replacement for Google search or Google on steroids. Before the advent of ChatGPT researchers especially, have been clamoring for better Google search with more contexts, intuitive and relevant feedbacks.

With the new ChatGPT (Plus) features introduction for examples web online search and plug-ins, ChatGPT has becoming a very powerful and viable better alternative to Google search.

"faithful recreation of human bullshitting" would be more accurate.