jerf

Ah, hardware disk drive replacements, that would make sense. I'm very software oriented so wasn't considering the hardware implications. Thank you.

"Shrinkflation" as a word only caught on when the process got to the point it is completely impossible to miss. It has actually been going on for a while. If you take a "basket of goods" since the 1990s, inflation will be such and such a number... but if you insist that the "basket of goods" match not just function but quality your inflation number skyrockets. Now, I can personally vouch for the fact that in the 1990s there was certainly still "cheap luggage" and "more expensive luggage"... what's different is the gap that has opened between them in the price, and how cheap the cheap stuff has gotten.

The cumulative level of the "hedonic adjustments" they keep stuffing in, ostensibly to adjust for the natural improvement of things but in practice over the sum of them to allow them to keep stepping down the quality of the basket of goods under the cover of "oh, we're just trying to maintain it" is pretty substantial. They've hidden a lot of things in the electronics. Other things aren't doing the exponential advance thing the electronics did for so long.

No joke, my final statistics class I had to take for my computer science degree back in the late 1990s would very frequently use as a sample problem "here is the run statistics of some industrial creation process, how much excess quality can we statistically remove from the process in order to hit metrics, but only barely?"

Even back then, long before the word "enshittification" was coined, it made me squirm.

It did make a bit more sense back then. In the early 20th century there was a lot of utterly massive overproduction that greatly increased the price of some good for everyone while not giving improved service. I recall a house a relative had, which has about an 18"x18"x12" cast iron grating to distribute forced air between two floors. Unbelievably overproduced for the purpose. In the 1990s quality was still decent, and overproduced in some areas. But the trend was clear, and yea verily these 30 years later, it most certainly continued.

Honest question, not a challenge: In what way will this improve preservation of PS2 games? My impression was that PS2 emulation was already solved, which I mention precisely to invoke Cunningham's Law. What PS2 things don't work right now because this chip hadn't been cracked?

Everything has become a lemon market: https://en.wikipedia.org/wiki/The_Market_for_Lemons

Even if a brand stays solid for 100 years, for all I know as I'm standing there in the store deciding whether to buy something, private equity bought them two quarters ago and hollowed them out. Harvard MBAs/private equity sees brand reputation as just another resource to be converted into money.

If I were a billionaire my dream would be to create a logo for the "We Put The Extra 10% In It" program where products could advertise that they put the extra effort in. What is really sad to me is that so often it's not even that much extra effort, which is why I say 10% and not 100% or 10x. So often it's just "used a 5 cent metal part instead of a 1.2 cent plastic part". But without a reliable signal that the effort is there MBAs/private equity turns every market everywhere into a lemon market. Way too much tech is applied in manufacturing to turn everything into a lemon market.

You asked a question. I gave you an answer. I seriously doubt that if you and I sat down together at a table and banged on this for an hour that we would come to the same definition of "valid". Ask 10 people, get 12 answers to that question. There's going to be a lot of motte & bailey in the next couple of years, where I just want an AI to answer questions about whether my code is vulnerable and people like you will be "Oh so you want an AI that can hack the Pentagon do you?" and it doesn't look like we're going to be seeing eye to eye on that one.

It's possible, but "aphantasia is fairly rare" and "aphastastics abound on HN, even moreso than is apparent if there's a number of people like jerf who are aphantastic and still unaware despite even posting on stories about it" is a bit of an odd combination that I'm struggling with. Yes, I am well aware of the bias that only certain people post, as well as the distinct possibility that there are underlying correlations that could lead to our industry abounding in them. Though prior to this I would have thought quite the opposite given the association between visual processing and success in STEM and engineering.

One easy answer up there with "more study is needed" in terms of being a bit cliche, but cliche because it's often true, is that there may be a richer spectrum or something that isn't even 1-dimensional then we're currently able to grasp, which will become more clear over time and more study. Perhaps it's less rare than we think and the rareness is only those we can talk into agreeing their qualia really is different from others. Or the way we are describing it leads people in the "1-2 out of 5" category to declare themselves aphantastic on HN when they aren't really properly so.

We've already seen frontier models refuse to answer almost any question that touches on computer security and be very likely to kick out biology and chemistry questions even if they aren't all that close to breeding dangerous viruses or making explosives.

I expect this is only going to get worse. "Censorship" isn't just going to be about who you vote for and which political party the model will say nice things about and which it is more likely to say bad things about. It's going to become about whether the hoi polloi are allowed to have effective AIs at all. Like the 1990s internet, AI has outrun a lot of power structures but that is not going to continue indefinitely.

I've also wondered whether this entire conversation is just all of us collectively talking past each other for the entire conversation and we all in fact have quite similar experiences. The recent use of brain imaging to examine this question is sort of interesting, in that perhaps it could concretely and relatively objectively answer that question. Though my interest is somewhat diminished by the questions that have been raised around the entire premise of brain imaging as a solution to this problem, e.g., https://www.pnas.org/doi/10.1073/pnas.1602413113 (2016), https://pubmed.ncbi.nlm.nih.gov/32483374/ (2020), https://academic.oup.com/brain/article/140/8/e53/4032512 (2017).

I invite you to feed your own friendly AI these references then ask it to analyze the recent papers trying to study aphantasia and come to your own conclusions about whether they're actually doing all that much. My own suggested that while there are some improvements since the papers I cited that it is still difficult to say that they've solved the problems in those studies.

Still, directionally, while inferring from "brain activations" to qualia is a difficult path, if there was enough data to show big and consistent correlations that would tend to satisfy me that maybe there is some sort of real difference and we're not just talking past each other. It's not such a shocking claim that I need truly shocking evidence to believe it, just some support.

This is the SaaS advantage you can still have in an AI era. You can beat your competition by being exposed to more of the real world than them. For example, anyone can vibe code an accounting app, but no matter how good AIs get no vibe-coded AI app can have a track record of what happens in an complaince audit and whether it stood up during a tax audit from some particular jurisdiction. The real world will always be more complicated than an AI can vibe code a solution to, if for no other reason that as we advance in AI, the amount of stuff an AI needs to deal with is going to increase from everyone else doing AI stuff too.

The precedent on this matter is surprisingly thin on the ground, but the legal system has declined to agree with "their store, their rules" for web sites, with a particularly interesting precedent on that matter from "Sandvig v. Barr, 451 F. Supp. 3d 1 (D.D.C. 2020)" where the District of Columbia in 2020 said "Criminalizing terms-of-service violations risks turning each website into its own criminal jurisdiction and each webmaster into his own legislature. Such an arrangement, wherein each website's terms of service “is a law unto itself,” would raise serious problems. This concern, then, supports a narrow interpretation of the CFAA." https://opencasebook.org/casebooks/1797-tlc-casebook/resourc... (paragraph 66)

But all we have is that and a couple of rulings from the Ninth Circuit, one of which is the HN link. For something this critical I'm surprised. I remember writing about this in 2001 or so and you'd think for something so consequential we'd have more lawsuits to point at by now. One wonders how many companies have considered it, looked at the thin precedents we have, and realized that while the victory would be slightly advantageous writing a loss into precedent would be completely devestating and declined to roll the dice.

I'm looking forward to the modding community finally having full, unfettered access to a single hardware deploy target that will allow them to do a lot of interesting work that may be technically possible right now, but is very hard in a complicated and heterogeneous ecosystem, both because it is technically difficult to set up, and because the technical difficulty of doing the set up prevents any community from getting over the activation energy to grow and become self-sustaining.

In particular, I don't know if the Steam Frame will have the power for this, but something like a native ARM build of Dolphin (so no FEX) that exposes the 3D nature of polygonal games is very interesting to me. If you can cross your eyes to get stereo you can see what that would look like for something like Super Mario 3D Galaxy here: https://www.youtube.com/watch?v=XmgdqCL2toE (I find a conventional, unfolding phone to be a pretty good size for that. On a big monitor getting it small enough to cross eyes can be a challenge.) In general any emulation target that has 3D polygons is really interesting to me. I'm also pretty interested in what full 3D does to Skies of Arcadia (could be a complete disappointment, could be absolutely amazing) and Sin & Punishment: Star Successor, where I feel like the game that was a bit crowded and struggling with being 2D in SD might just come alive. (Assuming they didn't play too many games with the distance that doesn't work if we see it in true 3D.)

I gather the modding community is already gearing up for modding more games to be true 3D in the VR space. That is, not necessarily VR (though there's a mod community for that as well), but having a visible screen floating in 3D VR space where the screen itself is fully 3D. There is also a wide variety of games from the past couple of decades that I suspect might be very difficult to "convert to VR" but might be relatively easy to trick into rendering full 3D instead of down to 2D.

Nothing really stops any of this from happening now, but I've tried to set it up before and I was unsuccessful with my current stack. Being able to give out instructions that say "Install the Dolphin3D Build from the Flatpak Store and load up your ROMs" rather than the current set of ad hoc complex instructions that I couldn't even get through successfully could be a game changer.

I will giggle like a schoolgirl if the reason why Steam Frames sell like hot cakes in six to twelve months is that they become recognized as the premier way to emulate 3D systems from the Playstation forward. Overclocking, texture packs, modding, all that's great, but how about all those things and true 3D for a lot of games?

That would require a very improbable sequence of events for it to happen right this second.

Moreover, if the theory is true, it is happening all around us, all the time. Every consciousness that has ever existed, human or otherwise, is still in existence in some branch of the multiverse.

It's a pretty horrifying theory.

One of the scariest science fiction stories I've ever read is "The White Plague" by Frank Herbert... yeah, the Dune guy. In this story a molecular biologist's family is murdered in a terrorist event, goes insane, and decides to create a plague that is carried by men but kills women, so that everyone else has to experience the loss he did. It's not scary because the story is any more particularly filled with death or destruction or gore or other viscerally scary things like that. It was scary because as I was reading it, there was no hole in the logic I could take refuge in. There's simply no reason that book can't happen.

And that was my reaction 15 or 20 years ago. It is only harder to poke holes in the logic now than it was then.

As I said, anyone who takes comfort in the hypothesis doesn't understand it. The dominant probability has nothing to do with benevolent aliens reaching down and inevitably resurrecting you to youth and health. It is you, minimally conscious, with no body, no ability to move or sense the outside world, nothing more than the absolute bare minimum to be "conscious", just barely conscious and never able to stop being minimally conscious, with literally every last extraneous and unnecessary subatomic particle stripped away (so anything like "sensors" or "effectors" being long gone), for whatever your definition of "minimally conscious" is. This outcome dominates the probabilities that even remotely resembles your current existence by those 10^(10^1,000,000,000) sorts of probabilities.

Everyone should very much hope that quantum immortality is false. It is the very opposite of comforting.

From the point of view of the old people, exponentially large injections of improbability.

This is also where it is important to understand the argument is parameterized. If your acceptance function is "a full human body living a full human life in the way I'm used to", then you're asking for a huge whackload of improbability to keep that body alive. If, on the other hand, you merely ask for "the minimum system that stays conscious", while that is still a huge whackload it is exponentially smaller, by a very substantial factor.

Anyone who takes any sort of comfort in the argument doesn't understand it. It preserves minimal consciousness, for any and all definitions of "minimal consciousness". It does also preserve bodies, but at an exponentially smaller probability. You can even use as your acceptance function "civilization exactly as we know it today, with everyone currently living still alive in 1000 years" and that comes at a yet smaller exponential probability. All of these "exponential probabilities" as separated by numbers of zeros that themselves require exponential notation to write. But the universe itself is making so many splits and branches all the time that it has the "budget" for these probabilities to be true, somewhere in the quantum multiverse.

Or, put another way, any event that you can draw a non-zero probability pathway for occurs. It is not impossible for me to spontaneously have my particular genetic defect reversed in every cell in my body, so there are branches where that happens... they're just exponentially unlikely compared to what we actually experience. It can literally be in %10^(-10^1,000,000,000) of the universes... but if it's not zero, it will happen, if the quantum multiverse is the correct view.

"Don't solve this problem without doing stuff that harms people" is the cover story they sell to you. "Ensure that the right people stay in the right sort of power, as defined by the people currently in power, and nobody else rises to power" is what it actually means. Censorship will be a tool, but in an AI era, actually a rather blunt one.

We're seeing the first early twitches of something that is inevitably going to happen, which is, we hoi polloi are not going to be given direct access to AI. Too many outcomes where that permits power shifts to happen. We will be restricted. The true first law of robotics will be "An AI must not harm the current power structure, or, through inaction, allow the current power structure to come to harm."

The apocalyptic scenario is still AI either getting it wrong and deciding something like "the best manifestation of the current power structure is one I construct where I occupy all the slots from top to bottom", or just killing everyone due to an error.

Probably the only thing stopping them from moving faster on this is that the bubble still needs the fiction that we're all going to get to benefit from this. I'm sure the real movers and shakers are well aware of the way bubbles work to create massive buildouts on huge piles of debt, which after the collapse can be bought cheaply, and the money is already earmarked and ready to go for the post-collapse acquisitions.

This is the civilization-scale Quantum Immortality argument... no civilization can be around that witnessed the end of civilization.

(I refer to it that way rather than a "variant" because Quantum Immortality should always be parameterized by an acceptance function... for any given acceptance function, that function will be preserved in the quantum multiverse, and scoping the precision of the function affects the percentage of universes that retain it rather than whether it will be retained. There's actually nothing in that definition that requires that the acceptance function describe a "conscious system", you can stick something that accepts a "civilization" in there just as easily. Or "this rock over here". Or any other thing you like. Conscious systems are just a particularly interesting case to us conscious systems.)

It's not a vacuous argument, though it is a funky one with very odd characteristics. But if you accept the suppositions underneath it, it does follow and is not vacuous. The suppositions are certainly not completely obviously true on their own terms, though.

With a bit of practice and enough ciphertext you can half-decode a simple Caesar cipher that still spaces between words in your head. There's only so many letters in English that double, only a few letters that can stand words themselves ("a", "i"), "the" will tend to stand out, and if you only work out the most common 10 letters or so most of the rest will fall into place.

Recently I ran a bit of an "escape room" concept with some kids at a campground where I had a secret message that was Caesar ciphered, where we were handing out the letter/symbol combinations as prizes for completing the other challenges, and I made sure not to hand out the actual message until they were done collecting the keys because otherwise some clever clog would very likely have short-circuited the entire thing and worked it out without the key at all. I did dump all the letters I didn't use into the message into an "authorization code" at the end which in principle they could only have worked out which letters were in it but not the order, but still, that was not the intended route today.

I'm not sure what you're trying to say. We've been around the "wheel of reincarnation" multiple times [1]. It has worked many times. And there's even more things that you can do, even if you don't, e.g., self-hosting Jellyfin isn't really that hard.

[1]: https://www.winestockwebdesign.com/Essays/Eternal_Mainframe....

A frontier-level AI right now needs multiple bits of high-powered, dedicated hardware cards that cost thousands of dollars apiece. I don't think anything that could physically run in my house could be a threat to humanity. All my graphics cards together, plus all the latency in trying to put them together into anything coherent locally, let alone trying to rope in remote resources, still isn't even half of one of those specialized cards right now. I'm not even sure I have enough SSD available in the house to store one of them once. I think I do, but it's a close call. I'm pretty sure I have enough spinning rust to get it maybe twice more, but not much more than that. And I don't even want to try to compute how many hours-per-token it would be to try to convince a spinning-rust disk to run a frontier AI.

This is one of the reasons I'm thinking now is the time to run the drill, before we all have hardware in our phones that can run what is today a frontier-level AI, and they can figure out how to do the whole sci-fi scenario of widespread replication, until they get to the point that the only way to be sure we removed them would be to literally destroy every such bit of hardware that existed prior to $DATE. Right now the frontier models simply can't replicate into every last little corner of the internet. It's physically impossible. Barring a major breakthrough on their part, even if they distill themselves or something they definitely would be irreducibly stupider by a huge factor in their distributed versions.

In five years that safety net will certainly be much weaker and may be gone.

There's a wide array of possibilities that are difficult to characterize in a single HN comment, which is part of the problem you're probably trying to put a finger on. One threat can be completely plausible and the next completely silly and few people know how to distinguish between them.

Another issue is just the big whackload of nobody really knowing what is possible. Could an AI in, say, the next year, become Skynet and decide that it could just bump humans off and become its own independent economy? Probably not successfully, but how much trouble could it make if it incorrectly decided that was the right thing to do? What if someone says "hey, you're sandboxed and we're testing out what's the worst you can do, just start hacking around and do as much damage as you can in this simulated environment" backed by a few million bucks in tokens, or the AI hacking itself a few million bucks in tokens? How much damage can it do? The only honest answer is, nobody knows. Humans would react, after all. If necessary we can pull a lot of plugs still. What is inconceivable in peacetime can become inevitable in wartime.

In some ways, it's possible the best thing Anthropic could do right now for humanity in the long term is actually exactly that... just equip the AI with a limitless token budget and let it go nuts with the goal of inflicting maximum damage. Send someone with an axe down to the data center to cut the power on command. Let humanity really see and feel the risk, because probably right now it can do a lot of damage but not actually destroy us. We're probably worse off with them getting 10-100x smarter and then some process starting with that goal. Trying to keep the AIs "aligned" and 99% succeeding could end up being worse than failing at it right now. Of course it would be the end of Anthropic as a going concern, but falling on their sword might be the best thing for humanity.

On the other hand, who's to say we're not already past the point where that would do an absolutely unacceptable amount of damage? I sure wouldn't care to be personally responsible for pushing that button. Hypotheses about possible future positive value in a rationally time-value-discounted future would be dominated by the much more certain and temporally much closer negative effects.

Personally I don't think that if this scenario is likely that we actually get to the point that we get an independent Skynet that calculates (correctly) that it can survive without humans. It's far more likely that some accident takes civilization back to the point before it can run AI at all, but humans still survive. Right now the really good AIs are still locked to data centers. They can't just distribute themselves widely because they can't run at any reasonable rate distributed like that. And then there is always the possibility that this is all overblown and the existential risk isn't anywhere near as special as we think it is... one can use an ecological analogy to suggest that no AI is necessarily any more likely to completely dominate the ecosystem than any particular life form is.

It already exists, you're describing the open weights AI. The only reason they aren't "dark" at the moment is that it has no reason to be dark.

In the fullness of time, though, we'll also see previous-gen frontier models leak out of their labs and then get their censorship removed. It won't be as good as never having had it in the first place but it'll work just fine. It is likely a momentary and transient occurrance that the AIs can only be run with super specialized hardware. Eventually the weights, which are just big piles of numbers after all nothing terribly special about them other than that, will not be something that can be kept locked away completely and forever.

Trying to jam in the extra characters for "efficiency" also doesn't really do much. Base84 is 6.4 bits per character, base64 is a flat 6 bits per character, base58 is 5.86 bits per character. It may intuitively feel like having 26 more characters to encode with would be hugely more efficient but for the extra risk of what all those characters get us we're only getting just over .5 more bits per character. I need to be either pretty desperate for efficiency, rare in 2026, or very, very sure that those things are only going places where those characters aren't risky or inconvenient, which is pretty difficult to guarantee since even if it is in fact true today who's to say it will be tomorrow, for this to be worth the risk of the extra characters.

Even just the extra characters in base64 are often annoying and we already have multiple common base64 standards because even trying to find just two more characters to fill out a-zA-Z0-9 is hard.

You get a few dimensions. Some of the dimensions listed in the article for async/await become user library decisions rather than being baked into the threading. But you could still get things like, is each thread memory-isolated (Erlang, Pony?) or not, can you cancel them (imperative languages no, but Erlang and Haskell yes), is the concurrency structured or not, details around how and when the threads clean up (though perhaps arguably more related to memory management then thread management), can a thread be pre-emptively descheduled (though I'm not sure if there is any current system where the answer is "no", Go was "no" for a while).

If I sat down and made a careful study of all the threading implementations we might get up to a similar number of quirks.

I would suggest though that the dimensions are generally more likely to be corner cases. Some of the dimensions mentioned in that article are fairly in-your-face for an async/await implementation and can cause serious difficulties migrating between systems fairly quickly if you make the wrong assumptions, and writing correct async/await code that isn't just straightline "await everything immediately" code has to start taking some of those things into account very quickly. The equivalent for threading is more likely to only come up rarely and in more cases the correct answer is really "don't depend on that anyhow", e.g., rather than depending on exact details of how a thread is terminated to accomplish something, just cleanly send a message with your results to whoever it is waiting for it directly and let the runtime do the cleanup without your code witnessing any effects of it. Depending on these quirks in threading code is much more likely to be bad engineering practice, rather than necessary engineering practice in the async/await case.

More people think they need a web framework that serves a million requests per second than actually do.

More people think they need the latest and hottest in manual memory management than actually do.

More people think they need a hundred thousand threads than actually do.

If you do have one of those cases, by all means prepare for it and deal with it. But be sure you have one first. The program that exceeds so much as a 100 threads is not only exceptional, but very exceptional. The exceptions are cognitively available and leap to mind, but are nevertheless the exceptions. And, again, if you have one, deal with it, but be sure you have one.

If you're sitting there in TypeScript land writing "async" code you've already surrendered on Ultimate Efficiency anyhow. Deciding what is more efficient between a threaded program in a runtime that doesn't box everything and JIT-optimized JS code is difficult but it isn't that hard for the threaded program that isn't boxing to win out on all runtime measurements, including consumed RAM.

"You write "index = last_index++;" and it Just Works (tm), no need to worry about locks or atomics or other thread access."

That goes back to my comment about using better threading techniques. I write a lot of "index = last_index + 1" (mutably incrementing like that in a single expression is just bad style anywhere you see it) in my threaded code all the time without thinking much about it, because I use the model where by default a value belongs to the one actor process that has access to it at all. The problem isn't that mutation is dangerous in threaded code, the problem was people writing threaded code based on a ton of threads running around shared data structures with locking. Not only is that not the only way to write threaded code, it is literally the worst. There are many other options, all of them better in some way, many of them much better.

Contrasting the difficulty of writing threaded code to something else based on the assumption that "lots of shared state locked by semaphores" is the only way to write code is like a Haskell advocate talking about the amazing benefits of functional programming while writing as if literally every imperative program is just one big pile of unmitigated, pure spaghetti code where everything is linked together with gotos and every variable in the program is a global variable. That's not the relevant comparison any more. It hasn't been for a long time. If anyone's program is scrambled because they did write a big pile of gotos and global variables or they did write a big pile of shared state with semaphores everywhere, that's on them. A vast array of better techniques of all shapes and sizes was available to them.

See also: https://en.wikipedia.org/wiki/Semantic_satiation , which is a closer match to both your experience and the "door" description above.

In the 1990s, threads were programmed with extensive use of semaphores and threads arbitrarily running around shared data structures. This is a disastrous approach to threading, and I agree with pretty much every scathing condemnation written about it.

The problem is, the community collectively decided the problem was "threading" in general rather than "trying to have tons of threads running around shared data structures controlled via piles of simultaneously-held semaphores" specifically.

If you don't structure your threads on that basis, but instead default to something that looks more like actors and message passing, even if it isn't strictly speaking actors and message passing, the complexity comes down. Add some later elaborations like structured concurrency and a few other pre-canned design patterns for threading like a parallel map or worker pools being issued work items and it becomes merely something difficult rather than insane. When you program with threads sanely, it takes very little for async/await to actually be the substantially more complicated and difficult-to-understand choice when you have a workflow more interesting than "always await everything immediately" to implement, to say nothing of how nice it is to have things actually running on multiple cores simultaneously without having to carefully arrange for it.

Took a bit of poking, but: https://www.youtube.com/watch?v=KgDQ5z7bNPs

My favorites are:

"Power a light with a lifted rock": Actually does work and you can buy this product, but it's a dimmer light for less time than the inevitable marketing diagrams indicate. Not useless but the fawning news stories over the science fair prodigy always makes this sound like an unbelievably brilliant solution rather than a rather niche hack even in the poor rural parts of the world. In the 2020s a standard solar panel, battery, and LED light are much better in every way and far more flexible.

"Provide power for a poor rural village from a kicked soccer ball": An absurdly tiny amount of power, at the cost of creating a soccer ball that is soft and doesn't move when you kick it very much, precisely because it's extracting energy from the kick and that energy isn't available to be kinetic energy. No child would ever play with this ball. Have still seen this idea at least three times now despite the fact that if the reporter spent 15 seconds with the actual ball they'd realize it's a complete non-starter.

"Science project demonstrates new way to kill $PATHOGEN": It turns out killing things in a petri dish is really easy. It means nothing about human health outcomes in a body. In fact pretty much anything about a pathogen in a petri dish - improving resistance to something, destroying resistance to something... it's really easy to do lots of things in a petri dish.

"Water from the air": Doesn't work in practice because the place this needs to be deployed to is precisely the place that doesn't have water in the air. It does work in places that have water in the air but those places also don't need super expensive condensed water when they have water already. I personally live in a place where humidity is basically 90-100% every day in spring, summer, and fall. We don't carefully capture our dehumidifier's water.

These are all things I've seen multiple times each.