nostrademons

People were saying it 12 years ago too.

Note that it may actually be true. One way to look at it is all successful organizations and communities are continually dying. In order for an organization to get critical mass and start attracting people, it has to be far above the mean in quality; the mean idea, after all, never gets realized. After that, everything is regression to the mean, i.e. downwards, until it ceases to exist.

Yeesh, they should've learned that you don't try to get cute in a legal document.

I remember asking about the Crockford License's clause "This software should be used for Good, not Evil" in my Noogler training. And then having Chris DiBona need to tell me "for the purposes of this particular license, Google is Evil". Because the problem is that "good" and "evil" could mean anything in a court of law; they are 100% up to the jury to decide. And so any sane firm that wants to limit its risk and stick to predictable areas of law will not use them.

Yes, it's pretty clear to me that ritual seppuku is not meant to be taken literally. Juries are instructed to consider the text of the contract carefully though: do you have confidence that 12 people chosen at random will reach the same interpretation you do?

We'll see.

Another way to put this is "Do you feel comfortable shorting NVDA tomorrow morning based on this news?" Because if Huang is removed it's very likely they will face challenges finding and reacting the next big use-case for GPUs, not to mention capitalizing on AI (which already is seeming shaky for other reasons). If you're completely confident that this will result in his removal, it seems like there's easy money to be made here.

Personally I don't feel comfortable enough to put my money where my mouth is. I'd give it even odds, but even odds, to me, is not enough to say it'll go one way or another.

40% is a pretty high percentage where they don't get removed.

Eh, NVDA could be in for some trouble if he's dismissed. He's very much a command & control CEO - his deputies are largely tasked with implementing his decisions, leaving no obvious successor - and the recent history of NVidia is one of Huang continually finding or creating new use-cases for its GPUs that has it keep attaining new heights even after the previous one crashes. There was a time, after the crypto crash of 2021, where NVDA was down about 66%; likewise after the crypto crash of 2018. If nobody finds the next great application of GPUs, we could see NVDA down 90% or more.

I mean, the last 3 paragraphs of NVidia's statement are:

> While the Board has no reason to believe that either national security or NVIDIA secrets were compromised, the Board's inquiry will examine all available information.

> Jensen fully supports the review and has committed to cooperating with it.

> The Board has confidence in Jensen and his leadership team and will provide further updates as appropriate.

My interpretation of this is simply this is what you do when you are the board of a public company and the U.S. government has accused your CEO of being the target of espionage. You need to get the facts, if only to cover your own asses. If it turns out the accusations are hogwash, then you have facts of your own that you can release to counter whatever hogwash the government puts out. If it turns out your CEO committed a crime or otherwise did something improper, then you need to dismiss the CEO to prevent your complicity in the crime. If it turns out there holes in your security protocols around top executives, you need to plug those holes.

But until you have the facts, coming from an independent investigation, neither they nor us have any clue what the right response is.

I think that endurance is an intrinsic limitation of drones, but I take the opposite conclusion as you. Instead of it meaning drones will never catch on, it'll mean that military missions and consequently the size and structure of the state will adjust their form to suit what drones are good at, rather than drones not catching on. In other words, we'll get drones, but we'll lose nation-states, rather than keeping nation-states but having them fail to adopt drones.

Drones' ability to absolutely dominate the battlefield inside of ~200km is just too powerful. You'll end up with small regional powers that are basically impregnable to invasion, but also can't project power because the range of drones is too short. City-states, basically. And because the defining characteristic of a state is its monopoly on physical force, the structure of the state will adapt to the dominant military technology of the time.

I do expect them to IPO in ~6 months - they were planning to IPO in 2026Q4 but they torpedoed those plans, and Sam Altman said it's off the table for the rest of 2026. The implication is that it's on the table for 2027Q1.

Interest rates are the big wildcard. If this is the start of a significant rising cycle, akin to 2022/2023, then the AI bubble is toast and all these companies are dead. If the Fed pauses in December/January there may be a window of opportunity in early 2027 and we'll see both OpenAI and Anthropic IPO and possibly jump.

Rising interest rates are a triple whammy to AI companies: 1) risk capital always flees first when rates go up, so the available capital dries up 2) they and their suppliers have borrowed significantly to fund datacenter expansion, so rising rates will directly increase their expenses and 3) AI is considered an investment for their customers, and corporate finance always pulls back on more speculative investments when rates go up, and so their customers will likely pull back and directly shrink their revenue.

It seems very much in the spirit of allowing concatenation and square roots...

So there's a common dichotomy in software engineering between the "big ball of mud" and "beautiful diamond" patterns. A big ball of mud is a software system that grows through accretion: it often has tangled flow control, data whales, lots of special-casing, lookup tables, and so on. It evolves, it isn't really designed. A beautiful diamond is a software system that is meticulously designed, consistent, oftentimes with tight interdependencies between how it works. Everything is efficient and easy to understand, but if you take one part out of it, it tends to fall apart. Common Lisp is a big ball of mud, Scheme is a beautiful diamond. C++ is a big ball of mud, K&R C is a beautiful diamond. Windows is a big ball of mud, BeOS is a beautiful diamond. Microsoft VSCode is a big ball of mud, Turbo Pascal is a beautiful diamond.

The big ball of mud structure is often taken to be an antipattern: at least, when I Googled this, Gemini told me it's a "haphazard, tangled anti-pattern" and gave me some advice about how to refactor such a software system into a beautiful diamond. But nearly all software systems that people actually use and maintain are big balls of mud. This is because the big ball of mud tolerates partial failure extremely well. It tolerates changing requirements very well. It tolerates parallel development very well. You can change part of it without throwing away the whole thing.

AI is a beautiful diamond. For the most part, if an AI model malfunctions, you have to blow it away and retrain it. Or I guess you can build layers to fix up the output case-by-case, but then you're really building a big ball of mud around the beautiful diamond, and obscuring it in the process.

That's a little worrisome when it comes to durability. Living, enduring systems usually end up becoming a patchwork of mud. Your ability to do this with LLMs are basically limited to LoRa patches and post-processing steps.

It already does this. I've found it'll suggest my kids' names frequently, after just 1-2 letters. It didn't do this before they were born. I assume it's training on my previous messages.

I was there for a couple Google Suggest code yellows around naughty or racist suggestions and it's basically an email going out that's like "The Chinese government is making us remove offensive suggestions from Suggest. We need any native Chinese speakers here Saturday to come up with a list of anything that might be considered offensive. Pizza and swag will be provided." Or it's reactive to PR flaps: "[why are black] autocompletes to [...people so dumb]. Can we please block this suggestion?"

The concern with features like these is that the suggestions are considered to be corporate speech that the company is endorsing, and so it can be held liable for any that break relevant laws. Internet companies are generally considered common carriers when they're just passing around stuff other people write, but when they are generating their own content through machine-learning algorithms they are held to be the actual responsible party.

I'm surprised we haven't seen more lawsuits and criminal prosecution against LLMs so far, though it probably helps that the current U.S. president is a combative asshole that prides himself on how much he can offend every group under the sun.

Technically not stealing, it's a legal workplace benefit. Possibly abuses the spirit of the policy and may get the policy changed if enough people do it, but he's only doing what he's entitled to.

And there's actually a good chance that PG&E is not being paid: most of the big tech companies have large solar installations covering the parking lot which power all the EV chargers, so they cut PG&E out of it entirely and produce their own power.

How do you determine the "correct" price? Something is worth what people will pay for it, no more and no less.

Markets have the advantage that many peoples' opinions are aggregated into the market price. This is far, far more accurate than having one person's opinion (an elected bureaucrat) set the price, particularly when they have no skin in the game.

If the IPO price is really 10x the public market price, that implies that the set of institutional investors collectively estimated the price 10x higher than the full set of retail investors. That's pretty unlikely: institutional investors usually do pretty extensive due diligence on IPOs, and you're already averaging out the statistical fluctuations by selling to a wide set of them. But if that does happen, it just means that institutions got fucked over to the benefit of retail (who gets a 90% discount on the same shares), which seems like something we should be celebrating.

You want at least one actual PowerWall so that your house doesn't go down when you take your truck on an errand. But yeah, if you're buying 4+ PowerWalls you're probably better off getting an EV with V2H capabilities instead.

I've got a friend who does it simply to fuck PG&E and avoid an electric bill. He has workplace charging, a Cybertruck, and a powerwall + Tesla solar setup at home. He drives his truck to work, fills up on electricity, drives it home at the end of the day, and uses it to power his house. Repeat the next day. The Cybertruck was cheaper than the 13 PowerWalls or whatever that's the equivalent battery capacity.

I think that major gateways like Google and Cloudflare and Akamai are doing things like TLS-fingerprinting and smearing.

I normally use a VPN to access the Internet. It's an ordinary consumer-grade VPN, and I use it for pretty innocuous things like watching the international versions of Netflix or avoiding being geographically-profiled and charged more when I book travel. The VPN itself, I'm sure, is being used for all sorts of naughty stuff: when I look it up in whatismyip, it comes up as VPN, "abuse detected", and "residential proxy detected". I expect to get blocked occasionally on it and just pause the VPN then.

What I found surprising is that pausing the VPN doesn't always clear the CAPTCHAs. The next request will also fail, sometimes. For a while I was wondering if I have a residential proxy on my network, but I went hunting with lsof and bandwidth monitoring and a list of all the open connections through my router, and it seems clear. Yet even with VPN disabled, I'll get CAPTCHAs from Google and Cloudflare.

I suspect they're using browser/TLS fingerprinting and then smearing the results of that across bad IPs. If you connect to them through a known-bad IP, they not only will CAPTCHA you, but they will record a fingerprint of any data they can. Then if you connect again with that fingerprint, regardless of the IP, they block that too.

Ironically, if you are actually malicious, you can get around this pretty easily. Because you'll know not to use a fingerprintable browser, or to rotate fingerprints. So you have one completely separate identity on a given IP (that is not yours), and if it gets blocked, you roll up another completely separate identity on a different IP. But this is too much work for an ordinary consumer who just wants to keep their IP private. So you end up with this excluded middle where the only rational strategies are to either do exactly what the website expects you to do (with no funny business, even if it's innocuous), or to go whole-hog as the aggressor where you're determined to skirt their policies no matter what, and if you're doing that you might as well amortize the effort by extracting everything of value you can from them.

It detects if your IP is a residential proxy, not if you are using a residential proxy. From the perspective of the outside Internet, they are the same thing, the only one who cares about the difference is you.

At least in imperative languages, a parser generator can execute arbitrary code, including side effects. There's nothing stopping that code from simply returning a data structure; indeed, most productions will have semantic actions that just say "return new Mul($1, $2)". However, the ability to add imperative side effects is what allows things like keeping a comment counter for nested comments; keeping a symbol table in case the same token might be a keyword or variable; handling indent levels in newline-sensitive languages; and so on.

Even in declarative languages like Haskell, your parser combinators will usually operate in a state monad so you can do things like this. True context-free grammars are rare; many real programmer convenience features have some level of context sensitivity, which is usually handled more easily by side effects in the semantic actions than by generalizing to a context-sensitive parser.

This is standard CS terminology - if you take a compiler course it'll be explained in the first couple lectures.

The ELI5 is that "semantic actions" are bits of code that a parser generator runs, and "production rules" are patterns that tell it when to run them.

Note that the terminology in this article is somewhat non-standard. I usually think of a handle as a pointer-to-a-pointer: it adds another level of indirection so that memory allocator can move objects in memory without needing knowledge of all outstanding pointers. IIRC the classic Mac Toolbox (~1990) used them extensively, and that concept seems to be what your Ph.D thesis is about.

The objects discussed in this article would probably be better called "descriptors". They're very much analogous to file descriptors in UNIX: the API returns an integer index into a table, which contains the actual data being referenced. The memory address of the table or its members remains opaque to calling code, and there is no way to access outside the bounds of the table.

> globally top-10 public company that boils down to a binary call option, and the probability underlying that option is inscrutable. That's never happened before

FWIW Bitcoin had the same dynamics. Its valuation between 2011-2021 basically derived from a superposition of two futures: one in which it took over the entire global economy and became the dominant reserve currency, and one in which it failed and its value went to zero. The middle-ground was effectively excluded because currencies have strong network effects and a currency that is not dominant within a market quickly ends up unacceptable to merchants and worthless.

It's still somewhat in that grey zone, used as a form of digital gold by people who believe that the dominant reserve currency (USD) is going to zero as well as a trading medium by rogue nations, but its performance has been lackluster since 2021. It's currently up about 20% since then.

Entirely possible that AI follows a similar trajectory. People assume that it's either going to go bust or take over the economy, but people are really good at overlooking excluded middle grounds. Personally I think it'll settle into a space where it's very effective as a coding assistant, effective at creating propaganda and fake forum posts, and has the effect of killing Google Search (but not really replacing it), but doesn't fulfill all the hype around completely replacing the economy.

Paper wealth unless you actually sell. I know several people who were paper decamillionaires in the dot-com boom and then their equity was worth literally zero a year later.

For that matter, that could easily happen to Anthropic millionaires. There's usually a 6-month lockup on employee sales after IPO. They are hoping the stock market (and specifically Anthropic stock) does not crash in the 6 months before they can sell, and this is far from a given in a rising interest rate environment with the potential for renewed hostilities against Iran. They also to be sensible enough to take profits off the table and diversify when they can, which not everybody is.

I wish we could talk more about Natural Stupidity, which is the flip side of Artificial Intelligence, the seeming desire of humans to show that they are dumber than a machine.

This is part of what they mean by girl names driving the shift. You have a traditional ethnic name that may be masculine or feminine and has a baseline low level of popularity. Then you have a popular female name that rapidly gains and loses its popularity. The effect on the mix is that it's shifting from masculine to feminine, but really you should look at it as a newly-popular female name being layered on top of the traditional name. And you can often see this in the timescale, with a clear pop-culture trigger leading to the rising girl name.

Looking at just part of the data, I can see this with Arya (Stark, Game of Thrones), Paris (Hilton), Stevie (Nicks, of Fleetwood Mac), Dakota (Fanning), and Tatum (O'neal), and the masculine trend also relates to the fading of generationally-popular names like Jean (Silent/Boomer), Pat (Boomer), Joan (Silent/Boomer), Courtney (Gen-X/Xennial), etc.

Having kids, a medical condition that doesn't kill you, or getting to old age is increasingly a luxury.

You're inventing a plausible-but-probably-irrelevant cause when a more likely cause is mentioned in the study: students were not using their phones a lot in school before the ban.

The narrative says that phone use was already banned during classroom instruction in many of the schools that later instituted school-wide bans. The impact on students would've been minimal because realistically the only time they would've been able to use phones beforehand was passing between classes or at lunch/recess.

Though this is still relevant to policy-makers who are thinking of instituting district-wide bans. It's sorta like all the heat around critical race theory: banning something that doesn't actually happen much in schools.

It's true that anyone with the key and an Internet connection can move money out of a crypto wallet at any time, but the point of the cold wallet is that they don't get the key because it's kept on a computer (or other storage medium) that is physically offline and disconnected from the Internet. Same reason government intelligence agencies use a SCIF for any important information work: you keep the information physically disconnected from the Internet.

I sometimes wonder if this is the true function of leaders in a democracy. To serve as a scapegoat so that the electorate has somebody to blame for all the unsolved problems in their life. It's psychologically much easier to believe that you're held down by the government and discharge responsibility for fixing your life to your government leaders than it is to take the long, challenging and often risky steps toward improving your actual life. Also, people who do take the latter path often end up with noticeably better lives over the years, and yet few people take that path.

Note that electing clowns to public office is not unique to NYC - we've had a reality TV star and an actor as POTUS (the latter also serving as governor of California), Toronto has elected a criminal as mayor, Minnesota got a pro-wrestler as governor, South Korea had multiple presidents resign in disgrace, and South America seems to have politicians ping-pong between prison and presidency.