marginalia_nu

That still just gets you autovectorization, and generally locks you out of the performance you could have with direct SIMD intrinsics.

Granted, the number of cases this distinction matters is relatively small, making a function faster only makes a program appreciably faster if that function is a bottleneck.

Yeah. It used to be unusable on mobile and great on desktop. But desktop has in my experience honestly been slipping pretty bad last few years. Maybe I live too far from the data center or something.

One weird tangentially related thing is checking whether a PR is merge:able after solving a conflict in this repo[1] for some reason takes several minutes. Maybe because there are 1000 commits in the same file. Doesn't seem UI related but weird regardless.

[1] https://github.com/MarginaliaSearch/submit-site-to-marginali...

The lack of reality is mostly from how much net wealth is a guessestimate. The actual realizable wealth is largely unknowable. There isn't enough price information to give a certain answer.

But sure, how do you propose to pay taxes with assets that can't be liquidated and may not even be possible to valuate?

Even if you somehow pay taxes in assets that can't be liquidated, now the government has the same problem instead. What is the government gonna do, pay its employees in unlisted stocks, yachts and famous paintings? How will it even know how much taxes it's gathered?

If the tax isn't isn't just satisfying some sense of petty envy, and the tax is intended to cover some budget deficit, I don't see how this would help.

If you have $2bn worth of the same listed stock and go sell half of those, now you have a net worth of $1400m because your gargantuan order drained the order depth, tanked the stock value and triggered a panicked selloff at the stock market which further drove down that stock's price.

You can't take net worth away because it's just an estimate of what someone is worth. It may eventually be possible to turned into dollars and cents without losing too much in the process, but almost universally it can't immediately be exchanged in such a fashion.

Even more so when we're talking shares in a company that is not yet public, e.g. a founder's shares. At that point the valuation is complete speculation, based on what the company may be worth in some hypothetical future IPO. There's no actual price discovery since there's no public trading of such shares.

Yeah it's a real pity what happened to TLJ. Both the first game and Dreamfall were fantastic in different ways. Though Chapters... yeah, could not get into that one at all.

Thanks, I guess :P

It's a well established term, and puzzle games like Myst are generally considered a subgrenre: https://en.wikipedia.org/wiki/Adventure_game

I think this was true, but I'll be damned if I wouldn't say The Drifter has taken the crown. It raises the narrative bar so much it's just mind boggling you can even tell a story with that much momentum and tension in an adventure game.

Overall we've been eating very well the last few years when it comes to new adventure games. In most of the 2010s new adventure games tended to not really be very good, but the stuff we've seen the last few years from the likes of Wadjet Eye and Grundislav are absolutely on par if not better than the highlights of the '90s.

Adventure games are games that are spiritual ancestors to the game Colossal Cave Adventure (often just called Adventure), which includes most point and clicks, but also the likes of Zork and Myst.

It's very often misused though, I think from people confusing the label as referring to games being about an adventure, rather than being like the game Adventure.

Yeah. Committing crimes via some sort of elaborate rube goldberg machine is not something that has been just invented in the 2020s with unclear legal precedent.

It's if nothing else a staple in the Arthur Conan Doyle tradition of mystery novels, and even then half the plots involved hiding who was culpable by using the machinations to create false alibis, with delayed murders and the illusion of a locked room as not even in the 19th century did anyone believe that the indirection itself would hold up as a defense.

If someone buries an anti-personnel mine in a public park and then argues that it acted on its own accord when it maimed a pedestrian, I doubt you'll find a court on the planet that would spend even a 4 seconds considering the culpability of the land mine itself.

When designing a standard like this you need to apply systems thinking.

The poor data quality is a problem for anyone who wants to use the tags, which has seen everyone almost universally reaching for other solutions. Search engines have preferred anchor tags, bookmarking solutions have applied user tagging.

With keyword tags it's been a vicious circle of poor data quality and neglect since day one. Even in documents from the early 1990s when people were really trying, the data quality is inconsistent at best.

Meta keyword tags have always sucked even without malice. Because they're invisible, authors frequently forget to update them, copying <head> boilerplate between documents.

I don't think I've encountered an external library that returned partially constructed objects returned nulls (at least not without a @Nullable). There are probably cases of this existing, but those types of libraries don't tend to see a lot of users.

Why don't you have any coding standards? If you're working with agents in particular, catching and enforcing this stuff should be easier than ever.

Have you tried not returning null or constructing incomplete objects?

News at 11.

Language that runs on the explicit design philosophy of letting other languages experiment first and then incorporating their lessons learned once the dust has settled is late to implement a feature.

Not at all a bad choice.

It's stable to the point of boring, and there's no shortage of people who know the language and can work with it, it's got best in class tooling, decades worth of libraries almost all very mature. Most of the language's issues are from legacy code bases coded in a style that isn't really relevant to a greenfield project.

So what I'm saying by saying that he fell out a window is that Putin and/or the FSB arranged his death. It's not a statement of the means of his death, but who arranged it.

I was alluding to how people who fall out of favor with Putin have a tendency to have mysterious fatal accidents, more than 10 of them falling out of windows.

Prigozhin falling out of a window was a not insignificant setback for their digital warfare capabilities.

Embedding the image would have ruined the comedic timing and made it less funny. Comedy is mostly about subverting expectations. If your eyes are drawn to the punchline before you read the set up it doesn't work.

> The current valuations are only inflated until they're not

You could say that about any historically inflated valuation all the way back to the tulip mania. Either the expected profit materializes or it doesn't.

> We're already seeing signs of this strategy from Open AI and Anthropic warning about the dangers of AI and the need for safety regulations

I would read this as a desire to pause training to be able to present a profit in anticipation of the IPO. The major AI labs mad scramble to IPO is if anything a sign that they aren't at all confident in the valuation. If they were they would be no hurry to cash out.

I asked Astra to explain this and it thought for a really long time and then came back with this:

https://www.marginalia.nu/junk/financing.png

Inflated valuations, mostly.

Yes[1]. Most people within the CAN-SPAM jurisdiction have their shit together well enough not get fined, but this does not seem to include many of the new wave of agentic spammers. Could be they're out of jurisdiction, or that they simply aren't aware or too naive and think this law isn't in effect.

[1] e.g. https://www.justice.gov/archives/opa/pr/permanent-injunction...

Big pivotal moment is probably the DMCA and the DRM it's enabled, it's really eroded the concept of ownership in both software and hardware (via firmware), to the point where we've completely lost all control over the devices we buy.

These days devices do what we ask when it benefits their real owners, and refuse to do what we ask of them when it doesn't.

It wasn't an overnight thing, but the temperature in the pot has been increasing for well over a two decades and it's getting more than a bit sweaty.

It's not like Doctorow, the EFF, and a bunch of others didn't warn us. They told us this would happen. Like exactly this.

Regardless who is suing the cost falls on the spammer. General consensus among these agent bros seem to be that these laws don't concern them, when what they are doing is exactly the sort of things these laws are targeting. Wheels of justice may grind slowly and this is the sort of stuff that doesn't hit you 5 minutes after you start spamming, but they do still grind.

Given CAN-SPAM fines are like $50k per violation, it feels like someone's about to find out.

Yeah I think almost anyone will agree that these matrix formulations are more confusing than helpful. The subject itself isn't really that complicated, but the jargon and notation makes it a lot less approachable than it is. Unwrapping the operations with explicit summation makes it a heck of a lot clearer what is happening.

There are places where matrix expressions are informative and helpful, but at least in the context of teaching machine learning, this isn't one of them.

Most larger businesses are kinda resistant to building stuff in house since it means that stuff has to be owned and maintained for as long as its needed (possibly decades) and you can't just rely on a SLA and the legal team to ensure things are working in perpetuity.

The danger of vibe coded SaaS:es isn't the customer building the thing themselves, but that the increased competition drives down the profit margin.