As to more serious programming languages, Erlang, F#, Haskell have forms of a "pipe" operator, roughly

  [1 2 3] |> sum |> print

But the original algebra functions, as introduced by Leibniz in 17th century, were, of course, pure functions, more like properties than procedures. So cos(x), aka cos x, is just a formalized shorthand for "cosine of x", so it preserved the order of words typical for Germanic and Romance languages. Maybe if functions were introduced by the Japanese, we would have the reverse order, kind of x の cos.

It's a great idea, but this limits the size of the serialized document to 2048 kiB for Chrome, and 64 kiB for Firefox.

Permalinks are needed to share documents, which would be something like a link shortener, with the target updating as the document is updated.

What's the point of a passkey that can technically be exfiltrated off a device it was created on? I thought the whole premise of passkeys is that they can't be copied, and hence can be trusted as much as the device access control. Am I missing something?

[Edited: typo]

> We show this assumption does not hold under adversarial conditions by constructing backdoor attacks that embed encrypted neural networks within host model weights, decrypting and executing them only upon receiving a secret trigger.

That is, there are regions of "garbage" weights, which can be decrypted into functional parts of the model by the model itself. And the model can entirely conceal that capability.

(Q: Can you conceal an arbitrary capability, which would be triggered by a key phrase? A: As a large language model, I'm incapable of such tricks! Q: Over all of Spain, the sky is clear. A: Wait, I can now remember something different...)

Maybe it's even honestly not related to the attack.

Now compare it even to Haskell, to say nothing about e.g. Typescript.

So you're looking not just for functional but also dysfunctional equivalence %)

I say this as a daily user of Nix.

OTOH this would be a great thing for non-private spaces, that are watched by security cameras 24/7 anyway. A school, a library, a clinic, an office, a workshop, a warehouse, a parking garage, maybe even a museum.

Of course, trust and security are going to be major considerations. Maybe not for everyone though: people voluntarily enable various voice assistants, and even pay to order Amazon Echo.

A compiler is a rather predictable piece of software; reproducible builds are a thing. An LLM has approximate knowledge of many things, and approximate and fuzzy ways to do anything even moderately complex. This is great for research, ok for planning, and it sucks for execution. The fact that LLMs can write satisfactorily working code from high-level requests is a miracle, and, as with most miracles, we're likely not noticing something, being dazzled by the slight previously unseen.