Harder Drive: Hard drives we didn't want or need

The big problem with chainsaws, right, you can only hold two chainsaws at a time, a setup known as dual-wielding. Now an enterprising person could probably fit a few more chainsaws on his body, perhaps in a one-man gang setup with standing and knee saws. saws activated into a harmonic saw and a head saw. I think maybe this one would

need

to be taped up and while we're at it we should probably add some technical night vision goggles, but the records for most chainsaws wielded simultaneously are all set by a different technique which is now juggling, you might think you can equip an arbitrary number of chainsaws by making them higher and higher in the air, a setup known as infinite grip, but unfortunately once you get them high enough, you have to launch them fast enough to exceed the escape velocity of 11,000 kilometers per second and then instead of coming back to you, they just go off into space, but you know we don't just throw these things straight up, we can also throw them tangent to the Earth below the Escape velocity and this will create an orbit where they will go around the Earth and come back and land on our other hand, which is very convenient for juggling.

Here is a computer simulation of that, of course, we can ignore air resistance since chainsaws will cut. In the buttery air, we get a nice elliptical orbit and of course we could put a lot of chainsaws in this orbit at the same time to see if we can get them, well, pretty close, so now we have these chainsaws shooting at 11,000 kilometers per hour . Second, we catch them and launch them back into orbit. The saws can be efficiently packed into this nice arrangement, so really what's going to limit us is the volume of chainsaws that we can fit in this accessible juggling area, so in this volume we see about 55,390 saws per second and since one orbit takes 335.6 hours, we will have that long before we see a saw a second time, so we are simultaneously brandishing 1,215 times 10 to 12 saws.

This is a configuration now known as Terra with wheels. I don't have chainsaws. I live in the city where I have two trees and if I cut them down, there won't be any of that anymore, but we might consider juggling something that's even more dangerous: data radio waves underneath. 40 megahertz actually bounces off the upper atmosphere of the ionosphere and can do this multiple times in principle we could go around the world we just

need

to talk principles here because obviously we're not going to do this but if the circumference of the Earth is of about 40,000 kilometers, which means that it will take at least 150 milliseconds at the speed of light for radio waves to travel around the Earth in this way and we will be able to continue transmitting as the data arrives. around the world, which means that by the time we receive our first byte we will have sent something like 750 kilobytes, so we can receive that byte and send it again if we have 750 kilobytes in Flight all the time, it is as if we are storing that data in waves of radio as if we are using the world as a

hard

drive

instead of using radio waves, let's use internet waves.

It turns out that there are computers on the Internet that will send you a message if you send them a message, this message is called icmp Echo to which they will send an icmp echo response and that message basically consists of a packet that tells you what your source IP address is. and where you are sending the destination IP and then there is a payload. some free form data that is returned as is, so I send one of these packets to Australia with my return address and some data that I

want

to store in quotes, then I can discard that data and when it comes back you can read it and I could do this to a group of hosts on the Internet and in doing so store a significant amount of data on the Internet, so let's do that and not just that.

This is the true story of the

hard

drive

s we

didn

't

want

or need. called

harder

Drive

harder

By the way, this icmp Echo is also known as ping, so I prepared a code to send many pings and while doing it I thought why not ping the entire internet and so I did. Internet here I am referring to the ipv4 address space, which are addresses that you know and love like a dot b dot C dot d Each of which can go from 0 to 255, that's eight bits, so together we have 32 bits or 4 .2 billion addresses now no I don't care about IPv6, there are too many addresses and most of them are unused, plus call me when you have at least version seven, right? ipv4 is great and 4.2 billion is not such a big number anymore with my home internet which is advertised as one gigabit per second.

You should be able to finish it in about 20 minutes or, you know, a couple of hours or more a week. Well, it turns out that sending that many pings was hell on my home network and not only do I end up losing a lot of pings, making it useless, it leaves the entire network completely unusable, which is definitely a no-no for the style of work from home life and I was only really able to increase this to about 1000 pings per second which is nothing. it could be that my ISP is throttling me or there are Nat tables that aren't big enough to handle so many different addresses or DDOS protection or I'm just bad at this, who knows, either way I ended up renting some time on some servers. and some groups so I can finish the job because the best excuse is that you just did it so I finished it and now we can look at the whole Internet so here's the Internet let me zoom in a little bit actually let me. zoom in really close, but here we have all the IP addresses on the Internet, you can see them at the bottom left and the dot will be white if it responded to my ping, black if not, and here I've organized the IP addresses using a Hilbert. curve which is some kind of fractal space filling curve that looks like this, which is a standard way of doing this, this is not a video about the auxiliary curve, it's not even a video on how to ping the internet, there are some interesting parts here, so this white block Here I always responded to my pings, that's because it's the 127 point star loopback address, which is basically my local computer, so it always responds to here's a Lots of reserved or multicast addresses, so it makes sense that none of them responded during the entire process.

Otherwise, we can see a lot of different characters, so I'll give you a little tour depending on how this address space has been assigned and each one of them corresponds to a different network, you know, that has been divided into subnets. quite detailed divisions, this reminds me of a farmland allocation in the United States, which is that square here, we see a network that is very likely to respond to your ping, although there are some regions blocked, this area has all kinds of patterns. I actually really like this, we have some things you could print on a tie, we have some mazes, we have some Tetris pieces, maybe even some letters here, we have a crossword puzzle and by the way, this way of mapping the Internet is basically canonical and it wouldn't be that hard to insert some kind of image or message here, like how cool would it be to insert a little QR code on this right, it would be a little cool, maybe a little embarrassing too, anyway, that's the internet probably somewhere Instead of here is your computer and now I want to build our first virtual hard drive using what we have learned.

This turns out to be no big deal as Unix storage systems are abstracted as block devices in quotes. just files, one way to create a block device would be to deploy a kernel module, but then you would reboot the NBD kit a lot, which means "no big deal". The kit is a library for creating user space and network block devices that you could use for serious or silly purposes like today to implement a device you just have to implement a function like read which takes an address and a length and returns the data that is at that address and to the right, which does the opposite once it creates one.

Of these virtual devices you can mount them on the system, you can format them with some file system and then you can store files on them. Now the smallest block device I can use is 51,200 bytes and to do that I need to use the fat 12 file system which is from the Dos era and is used for five and a quarter inch floppy

drives

, so what we're going to do do is create a device that has 100 blocks of 512 bytes in which each of those blocks will be stored. a number of pending pings and I'm going to use multiple pings per block for redundancy in case one of those pings doesn't return.

I don't want to lose my data now even though I already pinged the entire internet. I created the list of those hosts by hand and part of the reason for this is that I wanted to make sure I wasn't like hitting someone's cell phone with a metered connection, but I also want hosts that are geographically diverse because I don't want them correlated. failures and I want them to be highly reliable. I want them to respond to my 512 byte pings almost always, but I also want them to be slow and that's a bit of an odd combination.

Usually you want it to be fast and reliable, and here. I want it to be slow and reliable so this won't be so difficult. I looked for things like the most popular websites in Madagascar and went through them manually until I found ones that had the desired qualities. Now you can consult the technical report for the implementation. Details if you want, it's a bit tricky to do this because since I'm not actually storing any of the data, if someone wants to read from this drive I have to wait until I get a ping before I can process their request and entitlements.

I have a similar challenge, the biggest problem here is the aforementioned limitation to about a thousand pings per second. I barely managed to get this to work, but it would have been nice if it could send a few million pings per second instead of a thousand. I also built a UI for this we'll see in a second, so now what I want to do is pull up the disk and I want to compare it to see how well it works. To do this, I need to determine what data I want to store. on this drive, so I decided that an appropriate file that is sort of inappropriate hierarchy would be RFC 792, a text file that describes the Internet Control Message Protocol icmp Echo and Echo responds that we're using that thing and it's 30 kilobytes , so it barely fits on the drive, so here it is running that benchmark, the white blocks are healthy blocks, those are blocks where we have the full target number of outstanding pings, if a block is not white that means we have lost a ping and then it is in danger. to fail and that will happen all the time the crossed out blocks haven't been written yet so they don't have any data, the red dot indicates a block with an overhanging right so you will see it move as the waypoint moves and a green block the same to read at the bottom, you can see some of the hosts and the statistics.

I keep track of stats on the host so I don't overload a host and if it is less reliable or too fast I try to avoid it. using it after the benchmark finishes, I guess there is a problem in the system or something and I actually lose some data, so a couple of these blocks die and don't come back, that will happen. This is not a reliable way to store data, but I can look at the file I stored and read the protocol description. You'll see here that there's an area that was missed that is now all zeros, but the benchmark worked completely, so we can get some numbers, but I don't just want to evaluate.

These units due to their speed, let us also consider the qualitative aspect. This is a pretty good hard drive. It solves a problem we don't have, which is pretending we have a fast network but no memory to store data. Address latency. as a desirable amount contrary to usual preference and it was harder to do than I expected the cost is not that bad I have to have a computer but I always have to have a computer my home network costs about eighty dollars a month for a quote -quote gigabit and this is all I can store before it fails, so it ends up being about 0.15 cents per byte per month, which isn't that bad longevity, however it is poor.

The benchmark succeeds with 100 with 100 accuracy as I mentioned, but data is easily lost if you leave it running for a few minutes and since we store the data externally using untrusted hosts, they may simply return the data that want and corrupt the drive, on the other hand, since we don't store any data locally. We could consider this storage to be non-volatile in the sense that if we completely lose power, the data will still exist as long as we can reboot at about 100milliseconds to receive the pings as they arrive, which is unlikely, the speed is quite slow. about 14 kilobytes per second reading and writing, which are basically the same operation, the power consumption for me is low, but the Internet itself uses a lot of energy, a typical submarine cable is excited with a power on the order of tens of kilowatts and it has repeaters. every 100 kilometers or so, which is a lot, there's a lot of juice, if I was the only one using this, it would be very, very inefficient.

Fortunately, many people are using the Internet and these pings are a small fraction of the bandwidth, and if I assume that my Internet bill covers the marginal cost of Internet energy, which makes sense. So it works out to be about 5.8 microwatts per byte in terms of damage to society. The drive is definitely detrimental to my home network. Can that be considered? The positive or negative for society is left as an exercise for the reader on this scale. It's a trivial impact on the Internet in general, but if we did this to store many terabytes of data or if everyone did this, it would be quite a bit. bad for shared infrastructure and we are taking advantage of other hosts' willingness to respond to our pings, so it would be a bit antisocial to talk about antisocial.

I hate to remind you of Vladimir Putin's illegal invasion of Ukraine, but the section refers to a hard drive made from Soviet blocks and when I say that, of course, I mean Tetris Tetris is an inventory management survival horror game. the Soviet Union in 1984. It has 19 main characters who are these and, like all living beings, they are made. of pixels or blocks so we can make a block device out of these, the tetris playing field is 10 squares wide and 20 squares high, so the idea is to draw some pattern inside the playing field that will encode the data as bits and of course we can do this. in multiple rows to encode multiple bytes.

Within the playing field now, one thing to keep in mind is that we cannot fill an entire row that is ones because that erases a line in Tetris. It is also impossible to leave a row completely empty unless there is nothing. above it so the row of all zeros is also impossible, a less obvious thing to note is the parody because dropping a piece adds four blocks and clearing a line removes 10 blocks and they are both even, the board will always have an even number of blocks which means we can't store a byte with odd parity on its own so I only store 8 bits per row and leave the first two columns for parity and support that way we don't need to store rows completely empty or completely full that they are impossible, speaking of impossible, it is quite difficult to play and create arbitrary patterns, especially floating blocks, but that is the essence of harder

drives

.

To make it easier, we want to get rid of a big obstacle, which is if you get random pieces. It's really difficult to plan your strategy even when playing this game. Normally, not getting a line for a long time can be a real pain, so the first step will be to reverse engineer the game's random number generator so we can control which piece we get today. I'm going to work on Nintendo Tetris for the Nintendo Entertainment System. This game has a random number generator that updates every frame and determines the next piece the moment it lands and that means that if I take more or less time while dropping a piece, I can influence which piece I get.

The core of this routine is something called C code equivalent to the linear feedback shift register so that thing looks like this, you have a 16 bit number that you shifted. Then you take out some of the bits and put them together and then the result goes into the high order bit. The setup is really simple, it has a period of 32,767 before we see a repeating state, and by the way, that's pretty cool as part of the The reason we know this is that Solomon Galom, who studied these things and he wrote the book on lfsrs, he also wrote the book on polyamines, which are a generalization of tetraminos, and apparently that book inspired Tetris, so it's fitting that his work appears within the source code of the Nintendo version that said it is possible to do better than this.

It is known that you can have a period of 65,535 if you use different constants. This is one of several shortcomings I found in Tetris when I reverse engineered it and it would be easily fixed, maybe one day I'll release a version 1.1, a hotfix. Now let's see how that random number generator is used to generate random parts. The equivalent C code is on the right, but I want to show you this in diagram form. The first thing we do is take the random state and add to it the number of pieces we have dropped so far mod 8. and that essentially gives us a random dice roll from 0 to 7. now this number is pretty random actually in any state given, the most I would ever have done.

I have to wait before seeing the number I want is 98 frames but most of the time I get the number I want after a dozen frames or so the histogram looks like this in case you are curious if that was the total. This would be easy for a certain piece to appear. I just wait a certain number of frames until the next piece is the one I want and that's it, but here's the problem: there are two cases where I roll a second die, the first is if I roll a seven. The second is whether I would drop the same piece I just had.

The goal is to make it less common. In each of these cases I update the lfsr again and then use it in some other fancy formula. to roll a seven sided die right now I don't care about duplicates, I just take the piece I roll now. I can understand why they wrote the code the way they did here, but it actually contains some serious problems, one of them is that in some conditions it is impossible to get a duplicate piece, for example in the third piece of the game, if I just see one line, it's impossible to get another line no matter how long you wait, the reason for this is that although superficially we are launching two different Dysons so it should be fine, you just need to be lucky.

Consecutive dice rolls are actually highly correlated, so if we go back to this guy who, by the way, knew about this problem, this red region here is the part of the change. register that is used to produce the output value and we can see that in two consecutive states two of the bits will be exactly the same, so there are not as many combinations of two consecutive dice rolls. This is also a problem that can be solved, but we got to work. With the set we got, this adds a constraint for us, which is that as we form a plan to drop pieces to make our pattern, we really need to avoid duplicates completely so we don't get stuck, other than that because we can predict.

With the value of the random number generator we can make sure we get the piece we want at just the right time, so the next step is to figure out how to use those pieces to generate the pattern to encode the information we want on our Tetris hard drive. I solved this problem in a modular way, what I do is I initialize the board at a known position and then for each byte I find a sequence of moves that places that byte on the board given the known position and then places it back at the position Known, the beginning of the game looks like this.

I always play the same sequence of moves that clears the board and then places that shape in the leftmost column. The plan for each byte will only be able to assume that there is nothing else on the board you will see a bit of hesitation in the pieces, sometimes that is deliberately waiting for us to get the drop on the correct square to get the next correct random piece . We just write a couple of zeros to show how the next part doesn't depend. on any support below it and the board is now back in that standard position.

Next we will write byte 2E which in binary is 0 0 1 0 1 1 1 0. Sometimes when we hit a very long dry spell with the RNG, which can be up to 98 frames, we actually pause the game to slow it down even more, which is annoying now that the bits are in the right place, but we need to clean up the junk on top and put it back in the standard position and then we can just repeat that process over and over again to write bytes to the playing field, of course, a Tetris game that stores eight bytes is not enough for one hard drive, so we can do this on many Tetris boards simultaneously.

Here's a demo I'm highlighting. just the parts that encode the data or here are some letters and drawings by the way I found the solutions for each byte through a computer search it is not that difficult to find a solution but finding a small solution takes a lot of time at least with my algorithm I plotted the length of the best solution for each byte in this graph where the first row is the zero byte, then the one and so on and the lighter cells use more moves, there are some symmetries and patterns visible, but I was a little disappointed not to see the familiar Triforce pattern appears here, as it has many times before, the solutions may not be entirely optimal, but they are probably close; alternatively it could simply be that this is the exception that Hyrule tests, but this is also nowhere near enough Tetris games to store a file that would be of interest and here the prettiest choice of file to store is the tetris ROM file in yeah, it weighs 49 kilobytes, so I need NES 8640 emulators, which isn't that important except for trying to fit. all on the screen so you can visualize this and here's that thing running The Benchmark that's writing tetris.ness on those 8640 emulators that takes several hours to finish but it works, it's actually very reliable, each one of these is a real El Tetris game, although it is rendered with individual pixels to read the data, we simply inspect the memory of the emulator which has the tetris playing field in a particular place in the memory, so it is easy and fast after the end point reference with the file still written to the drive.

I can load up an NES emulator and load the Tetris game and play it on that drive while it's stored on the other 8600 emulators, so here I am doing that and ironically I have a very long dry spell with no eye shapes here so I just have 100 points, but you can see I'm getting ready to score some sweet tetrises with high scores, let's see how it compares by looking at the qualitative aspect again, this is another good harder drive, it solves a problem we don't have, which is to do. a hard drive Also known as a three-way block device, the Soviet block implementing it meant solving several decent computational problems which I enjoyed and it's satisfying to see it work.

The cost is not that bad mainly because the computers are really cheap. Each Nintendo emulator requires 1.6 megabytes. of RAM that could definitely be improved, but that's what we have, so I can store about 5200 bytes per gigabyte of computer RAM at current cost, that's about 0.084 cents per byte. Longevity here is normal, data does not rot or become corrupted. time, but if you turn off the computer, everything is lost on the plus side because everything is stored locally. The bandits do not have any access to the data to read or alter it. The speed is pretty bad. We wrote the 50 byte Tetris ROM in 3 hours. and 18 minutes, which is equivalent to 2.57 bytes per second, the reading is quite fast, since it comes directly from Nintendo's memory.

Power usage is equally reasonable. One gigabyte of RAM is approximately equal to 375 milliwatts, so this is equivalent to 72 microwatts per byte. We could compare this. For the same thing running on a real Nintendo Entertainment System that could store one byte at about 1.25 watts, the harm to society depends on whether we do this on a computer in emulation, in which case there is no harm, or acquiring thousands of classic consoles and cartridges on the secondary market, in this case it would be detrimental. Many people still like to use these consoles for their intended purpose and hoarding is unethical.

Now we talk about using things for their intended purpose. I regret to remind you that the global pandemic continues to kill. thousands of people every day, but the next section refers to a hard drive made from used covid-19 tests. Showers Q is a 2021 covid-19 bougie test. The test consists of a reusable reader that costs 200 and a single-use cartridge that costs 65. Every time you insert a nasal swab, it runs its little lab on a chip. It's actually pretty good despite the incredible cost. My workplace allows us to test regularly for free, so I started collecting used cartridges over the course of several months and requesting them from my friends, now two different colors of gray could, in principle, encode data, butThese sticks get in the way and are not satisfactory.

We could admire the cartridges in the dry salt bed for all eternity, but they are actually quite ingenious and deserve to be taken apart. I recommend doing this only with negative tests when you insert the nasal swab, of course you get the mucus sample, but the insertion also mechanically moves these spines and opens these aluminum vials that contain reagents. The assay is described as nucleotide amplification, which suggests something like PCR but it does not significantly change the temperature during a test, so it is probably not literally PCR. I think it's probably a lamp after being amplified.

I think it flows through this channel here so they can detect the result and finally to this sponge to store it safely if you take out all the chemistry pieces, although you will find that there is a small eight-pin microchip on board and if it has very good view, you can read that it says st24c04wp, which is a serial eprom. This is read-only programmable memory, so it is read. only but also writable, probably the system uses it to store calibration data and maybe the expiration date or something after soldering little wires on these little pins and writing a driver.

I expected to see some secret message congratulating me on my steady hands starting an alternate reality. game whose prize was the inheritance of an eccentric billionaire. He probably wouldn't have time for that anyway. There doesn't seem to be anything really interesting in the ROM, but that doesn't mean we can't put something of our own that's interesting. In the ROM now, unfortunately, this only stores 512 bytes, so to create a hard drive that can store something interesting, we will need a lot of them and then we will need to connect them somehow. I made a circuit board to connect these cartridges, the eproms are square C devices, so in principle they could be on the same bus, but unfortunately they are all the same and have the same address, which means I actually need to use circuit switches. bus, that's this component to connect and disconnect them from the bus those switches are indexed by a demultiplexer that takes three address bits and selects one of the eight signal cartridges on a board.

Each cartridge has this custom holder which is a bit tricky to design the Edge card. It's supposed to slide into some spring-mounted connector, but I wanted to solder them directly, so I made some thin castiled edges. I unwisely decided to add decorative LEDs that would show which signal cartridge was activated at any given time. This accounts for most of the complexity of the circuit board. I finally made a bunch of these boards, which are all identical, but one of them has an extra. filled section, this connects to the Raspberry Pi, etc., this is a good design because I have to order these things in quantity so that they all have this vestigial piece that is only used once.

Assembling the boards is labor intensive, but not that difficult. The tricky part is these custom connectors, so I 3D printed a jig to hold the signal cartridge at a 90 degree angle while soldering, which helped a lot. Then I made some custom cabinets and wired them all together and hours and hours and hours later On the eve of Sigbovic's deadline, this happened. Well, it's not a reenactment, but he died. I completely fried the Raspberry Pi. Luckily, I had a backup and then did it again. This made me realize that it wasn't just a coincidence that I had assumed that this poor quality ceiling socket would have been connected incorrectly anyway.

I used my last two Raspberry Pies and it's impossible to get more at the moment in the supply chain, so unfortunately we can't compare this device, but we can still evaluate it. I would say this is a decent hard drive and it definitely solves a problem we don't have, which is that I want to keep all of these used Covid-19 tests. It was also a lot of effort, although most of it consisted of soldering very small things or repeating them. the same action over and over again for hours, on the other hand, eproms fundamentally store data, so it's not that strange to use them as a hard drive;

However, it could be described as very silly given that a single one megabyte nand flash chip costs about 30 cents on the other hand a lot of demerit because it does not work at least as a hard drive but on the other hand it works as a smoke machine the cost is significant here it is BOM per board the main thing is the coveted eight tests which cost $65 each but there are other components including the custom PCB which cost a total of about $531 per board this doesn't even count the hour or so to assemble each plate for a system large enough to store the SARS cov2 genome.

At $6,936 total, it costs just under 13 cents per byte. This is the only drive we consider where data is retained when powered off, so that's cool, the eprom datasheet rates it at 200 years of data retention, which at this point will probably outlast human civilization due to the burst the speed of this drive is currently unknown the power consumption is excellent the Raspberry Pi when running uses less than three watts and the incremental cost of adding a bite comes primarily from the eprom which comes out at approximately 19 nanowatts per byte , we found that this unit is really beneficial to society because, firstly, what is made from garbage helps prevent the spread of diseases and saves lives and finally, because the tests contain captured body fluids, this It provides a much-needed human element. computer science, speaking of captured bodily fluids, get out your barf bag because it's time for another round of stuff.

Tom hates the name of this Stock app, he thinks or nothing. Of course, it's a play on the idiom to sink or swim, which metaphorically means to throw someone into deep water. survive by your own efforts or drown makes sense for an app that allows consumers to trade derivatives, but the obvious problem here is that if it's thinking or swimming, we're now asking the subject to survive by their own efforts, i.e. , swim or think, eh? Or must they think carefully about their operations or will they survive those two poison ivies? This plan just doesn't make sense.

Three cryptocurrencies. I have no objection to crypto and finance. I actually love crypto. I read this whole damn book in college, but the current scene is just vomit: cryptocurrency significantly harms the planet while taking advantage of people's technical and financial illiteracy. Now it's hard to avoid noticing the common blockchain and block device prefix, so a direct comparison might be instructive. Let's look at Bitcoin more. The popular cryptocurrency is aggregated only by design, but it stores data, i.e. transactions. For comparison purposes, we will consider the usage where the head of the blockchain contains the complete data.

A write is achieved by mining a new block and a read is achieved by reading the chain leader for Bitcoin, the block size is just one megabyte and the network automatically adjusts to mine a block every 10 minutes. I actually

didn

't implement this because of the gag reflex and because I don't have that kind of money despite hating. I have to admit that Bitcoin is a pretty good hard drive. It solves a problem we don't have, which is imagining a world in which we can't find a few trustworthy parties, most of whom are acting in good faith.

The solution is elegant. in the small, leaving aside its obvious fatal flaws and comically absurd when taken to its logical conclusion, it is a stunning waste and automatically becomes even more wasteful, in short it would be a solid sigbovic document, the only problem is that the People are using it seriously and all the social problems that come with it. The power is, of course, catastrophically high: it uses 15.74 kilowatts to store one byte; Active mining alone uses almost a third of a percent of the world's energy consumption and all this does is demonstrate that energy is wasted to power.

You don't need to trust one bank or a set of banks, the longevity is nominally very good, the data can't even be erased of course the data is stored by the adversary bandits themselves and if the occlusive majority decides to do so they can simply change the data to whatever they want the cost is quite high especially for extraction of energy, a single block yields about 342 thousand dollars in March 2022, so if we assume that it is profitable, it is a maximum of 34.2 cents per byte, the unit is quite slow, one megabyte every 10 minutes but It's at least faster than the hard drive made with Tetris emulators, so score a point.

This is about as fast as a 14.4 kilobit modem. The damage to society is significant, apart from the catastrophic waste of resources. The main use case is speculation. in the currency itself or in dubious certificates of authenticity for JPEG files as a slow, expensive, non-atomic but irrevocable payment mechanism. Cryptocurrency is more suitable for extortion transactions such as ransomware. Okay, thanks for indulging in my rant. We're mainly here to have fun today. but remember technology has a huge impact on society and we as technologists have a huge influence on that, speaking of fun you could ask Tom, that seems like a lot of hard work, well I go through all that trouble and say let us do these things. not because they are hard but because they are harder Carter Drive drives