What did they think Google would use, old VHS tapes?

Depends where. In the US, in the corporate world maybe. (I am in an European country, working in academia).
Here around LinkedIn is considered barely useful. Nobody would find weird if you don't have an account on that platform.
Countless local CV-hosting platforms seem to be more popular for job hunting.
As are also online portfolio on small webpages (github.io seem to be popular in my field of work).
Bluesky and even Mastodon seem more popular network in general in my milieu.

That seems to be very specific to some countries. I guess that the Zuck has managed to successfully becom "the web" in the US and some countries.
But very few businesses here around bother with facebook. Having a cheap static webpage (like some local hosting companies will host for free when you buy a domain through them) seem to be the most popular option.
Followed by listing on various business rating platforms.
If social network are involved, currently I am under the impression that a different Zuck's platform is more frequently used: instagram (mostly for showing pictures of the goods, specially for restaurants).

I've rarely heard classic social networks being used for sharing pictures with family and friends. The trust is extremely low in any of FB / intragram, etc.
For sharing for closed friends, chat groups seem way more popular, specially on platforms that (at least pretend to) implement end-2-end encryption.
WhatsApp used to a be a popular option and can still be find among older generations.
Signal is gaining traction specially among the younger (e.g.: all our PhD students use that for communication. WhatsApp is seen as an old people's chat network {insert here "in South Korea, only old people use" meme}).

Oh common, keep up with the trends:
"Not everybody will self-host a PixelFed instance."

Doesn't ANYBODY but me remember that "Napster" was actually RealNetworks? You know, the old Real.com that was the Internet's first scale, commercial streamer? Real became Rhapsody for several years. Rhapsody had no name recognition, so they bought the Napster name from it's owners... BEST BUY.

It gets weirder. Rhapsody had been Sonos' partner streaming service - and Rhapsody is also... I HEART RADIO. Now the whole Napster lot got dumped in the lap of venture capital vultures.

> it still doesn't seem like a meaningful improvement over KDE 3.5.

Have you tried LXQt?

On Debian just install it and uninstall connman and it's pretty good for most tasks, especially low-spec devices.

"So we're going to put out everything that we think is of the highest quality"

Let us know when you start.

One possible solution is to use RFC-1149...

And not even really that. At its core an LLM is a "plausible-sounding sentence generator".
It merely puts tokens together, given a context (the prompt, etc.) and given a statistical model (the distribution of tokens found in the corpus that the LLM was trained on).
It's like an insanely advance super-duper autocomplete on steroids (pun intended given the context).

If the model is rich enough the plausible-sounding sentence have a higher chance to be close to truth.
(Just like on a smartphone the autocomplete doesn't merely generate a gibberish string of random letters. With a good enough statistical model of the language it is targeting, it can auto-suggest keystrokes that form actual words, and those words are arranged in roughly correct sentences).

Yes absolutely. Specially given this part of the summary:

rather than genuine algorithmic reasoning

LLMs do not reason. LLMs cannot really reason. They can put plausible sounding words together that's about it.

There are some parlor tricks, like old school's "ask the chatbot to explain its answer" or the modern day approach of a "scratchpad", i.e., an internal intermediate storage where the chatbot can "write notes".
But that's again not real reasoning.
It's merely doing longer form of generation, generating an output that statistically looks like what would have been written by somebody writing an explanation or a reasoning.
But it's still merely generating plausible-sounding paragraphs of explanations/reasoning as long as those paragraphs fit the statistical distribution of tokens in the corpus the LLM was trained on.

Some neural nets have been good at solving sticky programming problems. Whether finding game cheats, doing voice recognition, modeling proteins, or other tasks humans haven't done well at.

But an LLM is more of an information retrieval tool, so tasking it with clever algorithm design is asking the wrong tool the wrong question.

Then there are the people who complete in programming challenges. In high school I would sometimes stay after to do the ACSL competition tests - no big deal, the school was a five minute walk, and it helped my buddies who wanted a high team score.

Then they implored me to go to DC on a trip for a national competition our score qualified us for. This seemed so bizzare to me as a fifteen year old kid - I could stay in a run-down motel and take tests this weekend or go camping in a state forest with friends. I let them down, in a way, but the ask was totally alien to me.

I have nothing at all against people who enjoy such things but it's a subset of the algorithm minds.

So we now have the results of some competitive coders vs. the wrong tool for the job.

OK, mildly interesting, but does it tell us much?

Are any of the profits going to Gila Monster conservation?

So every international traveler could be searched to see if their meds contain any patent violations and seized at the border?

This is entirely unworkable.

I wonder if they are doing this because they are losing revenue from no longer spying on Android users by using a sandbox hack.

Didn't they say some rogue VP set up his laptop to torrent all 72TB of Z-Library to feel o-llama?

I wish my laptop had that many drive bays!

Some people so want to believe that a useful information retrieval system is a superintelligence.

The rest of us aren't surprised that an interesting search engine isn't good at chess.

As can be seen in this table (from the FA), ASIF is considerably faster than the current UDRW.

I'm familiar with some organizations that have been feeding their Slack data into a RAG for employee queries.

They're going to be quite pissed if this has been shut down by Slack.