The code might be a little verbose which is tiresome for humans to read and follow. Structure and functions look idiomatic. It seems to be using xml parser idioms which makes it readable.
It could be doing double checks in both tokeniser and parser and things like that.
Actually looks like a good starting point and reference for someone working on xml parsers in rust.
Also knowing (archaic?) Scandinavian helps a little more.
"swa" is like a contraction of german "so wie". sindon is probably like german "sind": is/were.
soþ - sweet? gefeohte - past-tense born/nurtured/raised. ƿælfæst - wellfed. sƿylce - equivalent to modern "swole"? andƿlite - cognate with "anlete" which means face. ƿynsum - "finesome". searocræftum: specially-forceful (fantasy modern swedish cognate "särkraftigt"). "for þy" - since/because ("fördi"). forlætan: forgive.
ƿifode - wifed (strangely modern)
ofslean: probably closer to modern "avslå or "Abschlagen" than "slain". Defeat?
Ac - maybe like "ach"?
naƿiht: antonym to "evig"?
geƿitan - go/leave/escape/flee? (Scandinavian "vidd" means expansive landscape, cognate with "width" and "weit")
Nefne - negation of efne: "not even"?
stede - meaning is probably "farms" or "smallholdings"
gebunden - cognate with "bound", but the meaning is probably closer to "enserfed".
gefultumige - feels like past-tense of a verb that means "filled with"?
Squinting:
"And what she said was all sweet. I wifed her, and she was fully? beautiful wife, wise and wellfed . Not met I ever "swoler" woman. She was born so bold as any man, and though-whatis her face was fine and fair.
"Alas we never free were, since we never might from Wulfsfleet left, and never that Hlaford find and him defeat. That Hlaford had these places with such force bound, that no man may him forgive. We are here like birds in net, like fishes in weir.
"And we him secaþ git, both together, man and wife, through the dark strife this grim place. Whathere God us filled-with!"
He's not hired to code. He has taste for "what works" in these types of things. They want him to apply that taste - maybe making new services or fixing old.
> it costs almost nothing to build an app, it costs almost nothing to clone an app.
For the types of apps that AI can clone on its own, this has always been true. It's the eternal bookstore example, recipe collection, or my-dvd-collection app. The type of apps that Basic and Visual Basic were designed for.
If there was value in selling subscriptions to an app like this, it was probably coincidental.
I doubt they'd just add a UUID in a file header somewhere. If they uniquely modify the actual audio samples in a way that is inaudible during casual listening, that would be much harder to "diff", I think.
Can you guys catch up on the absolute basics of the last twenty years of audio watermarking research before continuing this conversation please dear god
Are you saying that two audio files each watermarked for a specific different user will not show a difference? Because then I am genuinely interested in how they achieved that feat
Watermarking might not be enough to prove that the person doing the distribution is the same one responsible of the leak. I fear at most it will be a contractual dispute between the person who received the watermarked file and the original distributor without the ability to easily link the overall counterfeiting charges.
But anyway, I don't think they really need to do that. They just need to shutdown any unauthorized distributors that make things too easy. As long as the friction introduced can convince people to pay a low subscription fee, they will be fine.
I think a latent embedding is almost equivalent to the article's hypernetwork, which I assume as y = (Wh + c)v + b, where h is a dataset-specific trainable vector. (The article uses multiple layers ...)
It could be doing double checks in both tokeniser and parser and things like that.
Actually looks like a good starting point and reference for someone working on xml parsers in rust.
reply