graceful6800

Be careful of who is looking down their nose at whom because my words come from having personally bailed out people who genuinely did not have the capacity to make sound financial judgements.

I'm not making judgments of the people you're speaking down about, I'm making judgements of a nation utterly failing to prepare its citizens for the financial realities of our world. It's not a personal failing for hundreds of millions of people to be financially illiterate, it's a failure of society and those elites who like to gaze down at the dirty peasants for not knowing better.

They were all lost when the library at Alexandria burned, so clearly we should ban libraries. Too much risk of losing unimaginably priceless works all gathered in one place like that

Saying "people should be free to get loans that fit their circumstances" implies a level of financial literacy that simply does not exist in this country.

You and I might be able to make educated judgments about a loan and its value/consequences, but the vast majority of Americans cannot. We simply don't teach this kind of thing in our schools, and every day innocent people get absolutely taken advantage of.

If the public were capable of accurately judging the utility/risk of a loan, then sure we can say it should be on the individual to make their own choices. But that is just not the reality we live in.

The fraud/not-fraud axis is wholly irrelevant once you consider that the profitable/not-profitable axis exists in a vacuum detached from all reality.

Obsessive is the right word. Over the weekend I had to stop it multiple times deep into a multi-hour long turn to ask what the hell it was doing. It was like a dog with a bone and would NOT let go of its current work to talk to me. I had to interrupt it three times with increasingly aggressive instructions to STOP and answer my questions before proceeding. In another session it straight up told me it was in the middle of debugging something important and to ask later.

I'm running an RTX 6000 Blackwell. It regularly spent over an hour per turn thinking. Every time I looked at it, the thinking trace seemed coherent, sensible, appropriate. But it could never settle on a solution.

Granted, I was trying to have it solve a hard problem that 5.6 Sol couldn't solve, but still.

Either way, I'm still impressed. It genuinely feels better than Sonnet 5

That's what I meant by carefully selected and ill-thought-out criterion.

They didn't set out with the goal of making gobs of money or subsuming the entire online population, and because those things didn't happen, success is tautologically ruled out, right? After all, as we all know, the only kind of success that is even theoretically possible is being the best and only player left alive in your field. Literally anything less is failure, unqualified and unequivocal.

Mastodon did what they set out to do and there's a vibrant community around it. That's a success, even if it doesn't meet narrow-minded silicon valley definitions of success.

Depends on how carefully selected and considered your criterion for success is.

It accomplished what it set out to do, which was not "make obscene piles of money" but since they didn't make obscene piles of money it probably is a failure anyway, right?

....in bed

That's the point of it though. It's designed to be gamified and addictive to coerce users into spending more time Engaging with the site.

Great that it doesn't work on you, but it works really well on a depressingly large number of people.

And the audience you want are users who will never, ever pay to get access. If GitHub starts throttling free users, they throttle the audience you're talking about far, far harder than you.

The audience will go away to somewhere you aren't. Then what?

Nobody, absolutely nobody wants or likes overlapping rainbow animated "ask AI" buttons plastered over every available inch of screen space. Nobody wants or likes useful features being removed and replaced with colorful bouncing "ask AI" buttons.

I run the same card and you're going to wish you'd gotten just a bit more VRAM :(

20GB is just barely not enough unless you want to live with 32k context, which just isn't enough.

The NVFP4 quant is completely broken, so I'm not shocked that other quants aren't fully there yet. Give it some time to cook.

I'm seeing the exact same number on my Blackwell box. MTP put it at almost 70 which is pretty decent

I feel like I'm the crazy one these days for using LLMs as tools. I don't write code anymore, but I still do the same level of engineering. I get fantastic results with minimal slop, and I feel it's because I still use my goddamn brain and engineering skills.

The way I use LLMs seems to be most analagous to the way vibecoder's LLMs use subagents. They exist to solve a bounded task or to write specific code in support of the engineering model in my head. I intentionally never did much research into how others were using LLMs before starting myself, and it seems like that was a good thing. It seems like the majority of people have given up on thought and just slam "do my job with no mistakes or hallucinations" into a prompt and then get confused and angry when it doesn't work. Or worse, don't check if it worked and ship anyway.

Americans will do and say literally anything to avoid admitting capitalism was a horrible, terrible mistake.

They literally stated in their first sentence that it was coding tasks.

Sure, just like age verification is what allows governments to protect children, right?

Netflix's moat is their CDN, which is fabulously expensive and complex hardware scattered across the entire planet.

AI isn't gonna help you magic a CDN into existence

That also is intentional. You join a server that suits your interests and follow people from other servers to broaden your feed. The goal is to form interlinked communities

More like "surgeon tried to cut out the cancer, missed, cut out liver instead. Patient still has cancer"

If you want to call "try multiple solutions to a hypothesis" brute force, then yes, literally all of engineering and science has been brute-forced since the beginning of history.

That's really what's going on here, I think. LLMs aren't that good at math or have deep mathematical insight. They just automate "try shit until something sticks" at scale, which, as it turns out, is a pretty decent way to turn out a bunch of math papers.

LLMs have enabled brute forcing of math in a way that classical computing somehow has not, which is fascinating in and of itself.

There's also a very old convention of using no units at all and assuming that "1" means uF.

That is one of the most insane sequence of words I've had the misfortune to read in quite a while.

Dunno, I'd mostly frame that as self defence from an increasingly hostile and authoritarian government and the actual, literal secret police.

Fighting back against people who straight up murder civilians in broad daylight is not something you can classify as extremism with any good faith.

I had 5.6 refuse to back up my android tablet the other day because it involved root access.

Outstanding technology, truly amazing what man has managed to create.

They're actually designed quite well for the intended purpose: mass surveillance and suppression of speech.

Don't ever for one second pretend this is about anything else.

There is no "market"

Using a duress PIN or other duress feature is now considered destroying evidence so YMMV