Adam: Hey, welcome to CoRecursive. I’m your host, Adam Gordon Bell. Today, we’re diving into a tale that’s filled with lies and greed and the very limits of technology.
So, Tom Perkins was a famous VC investor. The Perkins in Kleiner Perkins. They put money into Google, into Netscape, into Sun. But when Tom wrote his autobiography, he didn’t really talk about that very much. The main story that he included in the autobiography from his VC days is this crazy story. He had this friend, Roel Pieper, whose name I’m probably messing up.
( You’ll just have to get used to my inability to pronounce Dutch names. )
Roel Pieper calls him and says he’s got a new company and that they can compress videos like never seen before. Based on the description, Perkins is pretty sure that what he’s saying just isn’t possible. But he’s intrigued, so he flies to the Netherlands for a demo. Founding VC partners don’t usually do this. They don’t usually travel somewhere to see the demo. The demo usually comes to them. But he did it! So his friend introduced him to the brains behind the operation.
Jan Sloot, the demo was a bit of a mess, full of hiccups, but he did see it, live video compression, to an incredibly small file size, and he was satisfied. This invention could compress videos in ways that just maybe might break the rules of information theory. After the demo, Tom looked at the nervous man and said:
if this company works out, you’re gonna be the richest man in the world.
A statement like that, it could change someone’s destiny. But for Sloot? Destiny had other plans. While Roel and Perkins were working out the investment details, just a couple days later, Sloot was found dead. So was Sloot on the brink of a technological breakthrough? Or was it all just smoke and mirrors? And why did he die? And where’s this invention now?
Is his story one of groundbreaking innovation or is it a cautionary tale of greed and maybe even fraud? So to get to the bottom of this I’ve been sifting through documents and translated Dutch patents and I even spoke to the investigative reporter who broke this story in the Netherlands back when it happened.
So join me as we dig into this tangled tale, sort facts from fiction in the puzzling story of Jan Sloot and his mysterious coding system.
The Repair Shop
Picture this, it’s the mid 90s, a town south of Amsterdam, that I honestly can’t pronounce, and our story begins in a small TV repair shop. Not too many of them exist anymore, but you know, you walk in and the bell rings. Behind the counter stands Jan, 50 something, high forehead, deep set eyes, he’s tinkering with the back of a broken television set.
Sloot’s your classic tinkerer. As a kid, he took apart radios just to see how they worked. He’s been fixing TVs and VCRs for most of his adult life, but he’s got bigger ambitions. For years now, in his spare time, Sloot’s been working on an idea. Something that he calls Repabase. Imagine one giant computer database, and it’s got repair details for every electronic device in the world.
You type in the model number and the error code of your VCR, and boom, step by step instructions to fix it. No more manuals, no more guessing what part needs to be replaced. For Sloot, this is Repabase. It’s the holy grail of repair, of TV repair stores. It’s instant access to every solution. You see, Sloot is great at repairing things, but he’s kind of tired of the work.
He wanted to get others to do the repair work, but they don’t have the know how. How does he pass on his career’s worth of repair knowledge to a new hire? Well, Repabase is the answer. So he’s been compiling this massive digital encyclopedia, inputting everything he can. And of course, storing all this data takes up disk space, and disk space costs money.
Money that Sloot doesn’t really have. So he wishes for a way to shrink the data, to take a thousand page manual for a TV and crunch it into something much smaller. The latest compression schemes might shave off 25, maybe 30 percent, but Sloot wanted something revolutionary. He just didn’t know how to achieve it.
But then he had a breakthrough. He found a way to deduplicate the data. I think it worked like this. If every Philips manual has the same couple of starting pages, why not just store those once? Make the database not at the manual level, but at the page level. Encode it in that way and remove duplication.
And then maybe go even deeper. Only store lines once. If there’s repetitive lines, you can eliminate that duplication. Just like using a font can save you space compared to storing the pixels of each character, he finds with Repabase a way to store higher level information and therefore save space.
Jos Van Rossum
Sloot had a problem though. He’d made mistakes. He’d filed for bankruptcy more than once. He’d burned bridges with creditors and banks. No bank wanted to touch him. He barely had two nickels to rub together. How was he gonna develop his world changing invention? Enter Jos van Rossum, a businessman who spots potential but also sees serious flaws in how Sloot runs things.
Van Rossum makes him a deal, invests the equivalent of half a million dollars into Sloot’s company, but it comes with conditions. Van Rossum takes over, reorganizing the business into three plans. Plan A, Plan B, Plan C. The workshop repair business is plan A. It’s the cash flow. It’s where the money is coming from.
Plan B is the Repabase system Sloot has been developing. This database to streamline repairs. Maybe this could make real money. It could be licensed out or sold. I’m not sure. And plan C, plan C is this compression scheme that Sloot’s come up with. Maybe it’s useful outside of the repair store.
They seem like an odd couple, an eccentric inventor and tinkerer and this shrewd investor. But maybe they’re on to something, right?
But soon, Van Rossum grows frustrated with the relationship. The Repabase system isn’t ready, despite months of work and capital sunk into it. The workshop can’t even use it for its own people to tell them how to repair things. And plus, Sloot’s distracted. He’s spending more and more time on his compression technology than on getting Repabase off the ground or doing the TV repairs.
So Van Rossum had it. If Sloot wouldn’t deliver on the Repabase system, maybe his compression invention could be worth something. So he set up a demo, a high stakes demo with Philips, the tech giant. It was a make or break opportunity for Sloot to show what his technology could do. So with demonstration day just a month away, Sloot was desperate.
He scrambled, analyzing his notes and his experiments. He needed a further breakthrough. He needed some demo to impress the Philips team.
The First Demo
Then it’s demo day. At the Philips Research Laboratory, an hour south. He meets the two engineers, they’re there to see his invention, and they’re skeptical, but curious.
So Sloot hooks up his laptop, and he takes out a smart card. You know, the little chip on your Visa card? It can hold data, but only kilobytes of data. Not enough data to hold a modern high def movie, but also not enough data to hold a standard definition movie from 1997. But yet, he plugs it into a card reader, and the video appears.
A dancing man and woman fill the screen, twirling to music. It’s a nice demo, the scientists say, but how does it work? How did you do that? What sort of compression ratio did you achieve?
Sloot, Hems and Haws, he won’t answer. He says it’s a trade secret. But here, I’ll show you again. The Philips guys exchange glances with each other as Sloot repeats his little movie trick.
When he’s done, they break out the calculators. The math doesn’t add up, they say, the card’s capacity is nowhere near enough to hold the video Sloot is claiming is on it. His results are impossible, they say, and an awkward silence fills the room. So ended the first private demonstration of Jan Sloot’s radical data compression technology. If that is, it was ever real.
Sloot leaves Philips empty handed. And Van Rossum decides that maybe he’s made a mistake. He’s been working with Sloot since ‘95. They got a patent in the Netherlands on some of his ideas. But now it’s 1997 and it’s two years later. He’s got nothing to show. Maybe he’ll get his money back eventually, but he’s got other things to worry about. And so he moves on.
Sloot’s not done though. A year later, he’s got new investors. And they’re not interested in his TV repair store or Repabase. They’re hooked on this video demo. If he can compress videos to a kilobyte in size, that is going to change stuff, but they’re out of their depth with this compression and video tech. They need to vet what he’s got. And so they bring in experts.
So, in January of ‘99, Sloot gives a private demonstration of this device again, to a small group that includes experts from Oracle. Sloot claimed in this meeting that he could compress a full length movie down to one kilobyte encryption key, as he called it.
To put that in perspective, a typical two hour movie is around four gigabytes. At DVD quality. Sloot was claiming compression on the order of 4 million to 1. It was a bold claim. And the problem with experts is, they might be aware of how unlikely that is. Although on Sloot’s side, something you should know is that he never claimed this was a compression scheme.
In fact, he angrily insisted that this was nothing of the sort. The one kilobyte compressed movie sounds impossible, and Sloot didn’t argue with that. He argued that the one kilobyte was a key. The encryption key that looked up the movie. He said the whole principle was very simple, but if you asked him more about it, he’d get upset.
He’d claim it was so simple that saying more might give it away. It’s like you’re spying on him. How dare you ask him questions when he’s showing you it in front of your own eyes. Sloot was nervous, quick to anger. If you question his life work, he just might bail and not show you the demo. But this left his investors in a strange place, right?
They knew they had something special. They’d seen the demos of it. They knew he was brilliant. But if you throw him in a room with R& D people from Oracle or Philips and they start asking him questions, Sloot would get paranoid that they were gonna steal his idea. He might shut the demo down, he might not answer any questions.
But that day, at that lawyer’s office, Sloot did something remarkable. After everybody piled in and the lawyer made them all sign NDAs, they were brought into a room to see Sloot’s system. So it has a normal looking laptop running windows 95. He has a card reader that’s external to the laptop. It’s like a, you know, like a five and a quarter floppy drive, but instead of those big floppies, it takes in a smart card, like a credit card.
And then he has this laptop hooked up to a large external CRT monitor. The experts watched, he put the smart card into the reader on his laptop. pressed a few keystrokes, and all of a sudden, there’s a sharp, full screen video on his monitor. One of the experts there was Dick Vesters, a database programmer who was very interested in operating systems.
Dick knew that a Pentium 1 laptop running Windows 95 couldn’t play a flawless full screen video. He just didn’t think it had the hardware and software available. But there it was, and even more so, the laptop’s hard drive light stayed off. Dick took notice of this. The hard drive was idle. Even while the movie played.
Vesters racked his brain trying to explain what he was seeing. Either the light was broken, or the video data was streaming entirely from that smart card. To dick just the video playback at that speed, at that fidelity, at that performance, never mind the quote unquote encryption, just that display was a big deal.
The speed at which that happened was astonishing. I have never seen it anywhere since. Even with modern DVD players, there is some delay when fast forwarding and rewinding. The data stream could therefore be sent forward or backward in one way or another very quick. Moreover, he flipped back and forth between two films, and that meant that everything that I’d already seen must have been multiplied by a factor of two.
Dick got excited, and he started probing Sloot:
When I asked him whether he had indeed developed his own operating system with which the computer communicated to his card reader, among other things, he agreed. He said that was one of the issues he had struggled with the longest. I asked him if he used existing algorithms.
If he was really able to achieve that reduction factor, I figured he must have completely abandoned what we know. He must have used algorithms that store information in a different way than we’ve done so far. According to Sloot, that was indeed the case.
After the demo, Vesters and the other experts were dumbfounded. They knew Sloot had shown them something highly unusual. Even if the claims about compression seemed impossible, this was something. When asked for the verdict, all Vesters could say is:
This is incredible. This is incredible.
In later interviews, Vester would explain how Sloot’s demo just defied logic. Fullscreen video, idle hard drive, instant rewinding, none of that should have been possible. Yet Vester’s also believed that if technology like that really worked, if it worked as promised, it would revolutionize computing. I assume this is exactly what the investors wanted to hear. The tech had been vetted.
If they could successfully pull off demos like this, they could find the right company to license this code, and they could all get rich. In the Netherlands, the best company to license this is very clear. Unfortunately, Sloot had already burned that bridge. That company was Philips. They are based in the Netherlands, and they have a consumer electronics wing and a semiconductor division.
This is perfect for them. If the tech is going to change the world, it needs a company like Philips to see that through. But Sloot was already suspicious. He didn’t like the questions Philips had asked. And he told everyone that they would just steal his invention, to bury it, to prevent it from disrupting the industry.
But the investors persisted, and they convinced him.
So that in February, one month later, Sloot got a high stakes meeting with Philips. The CEO of the Philips Semiconductor Division, Roel Pieper, was there, along with engineers from NatLab, Philips elite research arm. Philips hosted them in an elegant 28th floor boardroom with sweeping views of the city.
Sloot set up his equipment again, including the laptop with the custom software and the card reader and a large 24 inch monitor that he always insisted on bringing. His paranoia was there too as he voiced concerns that the room was probably bugged. They were probably listening in, capturing the radio waves.
The demo had improved as well. I mean, the data I found was a little bit inconsistent, but it seemed like by this time, the laptop and the card reader were no longer connected. The playback device was the card reader, and it could now be hooked up directly to a monitor. This is a much more amazing demo, because there can be no doubt that the data is coming from the playback device.
The laptop is now out of the loop. Sloot gave a big speech:
After years of experimenting, I’ve developed a method with which I can put all types of data on a chip, using existing techniques, and without using any compression. And then play them back without loss of quality. With my method, I can generate a keycode for any film or program that is no larger than one kilobyte.
Then Sloot pulled out a white chip card, basically a blank smart card.
There are 16 films on here. I could have put more. The keycodes for these films are presented to a program in the playback box that is on a number of chips. There’s solid state, there’s no rotating parts involved, there’s no hard drive. The program decrypts the keys at an extreme high speed into exactly the same information that was previously encrypted.
Then Sloot inserted one of his smart cards into the slot. Suddenly a flawless full screen digital video appeared on the monitor. The engineers and executives, they could only gape in astonishment.
The semiconductor CEO, Pieper, peppered Sloot with skeptical questions. But Sloot firmly reiterated that all the video was streaming from that tiny chip. No other storage or memory was being used. Then Sloot showed that he could instantly rewind and fast forward the movie, and even switch between different films.
This sort of performance was simply impossible with 1990s technology, yet here it was, happening right in front of them. And Pieper was blown away, and he believed everything he saw. The engineers, they weren’t buying it. One started asking tough technical questions, and Sloot got defensive, and he blew up.
Pieper, sensing the tension, stepped in and ended the conversation. And as he left, Pieper told Sloot, “You did well, old man.” After they had cleared out, Pieper told the crowd, “This could be a computing revolution.” But the Philips engineers weren’t convinced it was a revolution, and there’s a reason for that.
How Compression Works
Compression is a well-studied thing. You can shrink data, but only so much. Imagine a packed suitcase, like a large one that’s pretty jam-packed. Now, if I told you I could take the contents of that suitcase and fit it into a smaller suitcase, like a carry-on, you might believe me. Right, I could fold things tightly, I could squeeze out the extra air, I could pressure pack the clothes into smaller bags and then put those inside the suitcase, and I could get it down to a smaller size.
But now imagine trying to pack your suitcase into a thimble. Clearly it wouldn’t work. Even if I shredded all your clothes into a fine powder and removed all the air, it would just be too much to fit in a thimble. That’s what it’s like to compress a movie down to a kilobyte. It’s like claiming that you can compress the contents of a suitcase into a thimble, or even more.
Sloot has sort of an answer for that, and I’ll get to that soon enough. But first, imagine you’re trying to make a big set of information smaller. Like you’re writing a compression algorithm yourself. How do you know if you’ve made something as small as it can get? There, there’s actually a theory for this.
Introduced by a man named Andrey Kolmogorov. He said that the complexity of a piece of information is the length of the shortest recipe that you can use to describe it. So let’s say you have a sequence of numbers and it’s 1, 2, 3, 4, 5, 1, 2, 3, 4, 5. If you notice, one way to represent that data is just as 1, 2, 3, 4, 5, 1, 2, 3, 4, 5, but a way to compress it is just to notice that there’s repetition in it, right?
It’s repeating twice. So instead of the whole number, you just say 1, 2, 3, 4, 5, times 2. That’s shorter, right? You’ve dropped off a bunch of characters. That’s compression. Decompression will do that multiplication back out into the bigger number. Can we make it even shorter than that? Maybe, maybe we can say 1-5 * 2.
And we’re still a bit shorter, but that’s probably about it. That’s probably about the smallest you can get and still get all the data back out. That’s Kolmogorov complexity. It’s like the packing thing. A t-shirt can be folded and compressed into a smaller and smaller space, but a brick cannot.
Similarly, some information can be compressed a lot, like that repeated 1 to 5. Or imagine if our numbers were just all zeros. That’s like a movie that’s just a black screen, while other movies are colorful and action-packed. And they just have more essential details, and because of that, they can’t be compressed as much. In essence, this Kolmogorov complexity is the limit on how small you can compress something. And it actually varies depending on what the data is.
So I assume the Philips engineers had discussions just like that among themselves, and so they told Sloot’s team they needed more proof. The Philips CEO of semiconductors, Roel, he was sold, and he was no lightweight, he was a senior executive, one of Europe’s largest tech conglomerates.
Philips had over 30 billion in annual revenues and 200, 000 employees worldwide. And as the CEO of the semiconductor division, he had a lot of clout. He could greenlight massive projects with just his signature. The point is this was not just some impressionable investor being taken in. By Sloot’s claims, Pieper understood technology as well as most tech executives.
And yet, something about what he saw defied his notions of what was possible. Roel started corresponding, eagerly, with Sloot’s team. He hinted at them that he wanted to leave Philips. And that he wanted to join their team. This is frankly insane, right? He wants to leave a CEO position at a 6 billion dollar company to join this TV repairman genius and his quirky team of investors.
He obviously sees something, right? He had been the CEO of Tandem Computers before being at Philips. He’s the type of person that probably gets paid huge sums of money just to sit on the boards of very large tech companies. And he wants to leave it behind for this magic device. Sloot was suspicious. But they did have meetings, and Sloot provided some glimpses and some high-level overviews of how this technology worked.
He resisted efforts to get deeper than that, because he claimed he was nervous that his secrets would be stolen. He said that his key-based encoding system was simple and it would be easy to steal. I don’t know, maybe he had other reasons for paranoia. Anyways, despite Sloot’s extreme protectiveness, Pieper plowed ahead with big plans.
In emails with them, he talked about going into hiding, drawing up these big plans before storming the market with this tech so fast that rivals can’t interfere. He spoke of spinning up an unstoppable network of companies to license the technology out to the world. In other words, he was excited, right?
He recommended linking up with some Silicon Valley VC folks, Kleiner Perkins, he had connections there. He wanted to assemble a world-class team and he wanted to commercialize this breakthrough across the world. For Jan Sloot though, Pieper’s interest was a vindication of all his work, of toiling in obscurity.
But it also did raise the stakes. Can this groundbreaking compression technology really hold up to scrutiny when we start pushing it out into the world? Or will it fall apart? Is there a hidden glitch in the technology that no one’s spotted yet? Who knows? People who trusted Sloot’s invention, they believed it was going to revolutionize the world. But they’re also getting a little wary of Sloot’s surface-level explanations.
Side Note: Shared Dictionary Compression
Okay, small tangent, speaking of groundbreaking inventions and the pursuit of revolutionary technology, one time I stumbled upon what I thought seemed like a groundbreaking idea in compression. And it was a moment that kind of reminds me of Sloot’s journey, filled with excitement.
I had heard in a math class that the digits of pi contain every possible sequence of finite numbers. This is a conjecture, it’s yet to be proven. But according to this conjecture, my phone number is somewhere in the digits of pi. If you convert this audio file to a large decimal number, it’s in there too, which is just a wild thought.
But mathematicians believe that that’s likely the case. If pi is normal, its decimal expansion includes every possible finite sequence of digits. I don’t understand why, but when I heard this I thought pi could be used as a shared dictionary for compression. Imagine I want to send you a part of this audio file, and we both have giant hard drives filled with the digits of pi.
Well then I could just send you the offset into pi instead of the file. That’s why it’s called shared dictionary compression when we both have the same data on either side. So we don’t have to transfer things, we just have to say where it is. It’s easy to see the potential of this when considering the Kolmogorov complexity of simple cases.
If I need to send you a 100 digit number, I don’t have to send you all the digits, I just have to send you the offset of where that 100 digit number is found. This feels a lot like what Jan is proposing. He insisted the card didn’t contain the compressed movie, it just contained a key for retrieving it.
He insisted it wasn’t compression, but encryption. I don’t know why he called it encryption, but it seems like what he was talking about was shared dictionary compression. Anyhow, Roel Pieper joined the company, left his Philips CEO position, and joined them. And in his position as frontman, he started explaining the invention.
Sloot makes an end run around conventional information transmission. Think for a minute. If I want to send you an image of the Mona Lisa, I could scan a photograph and send the bits in a stream to you, and you could reassemble them into the picture. That would take a lot of bits, and it would take a lot of time.
Or I could just tell you, look at page 75 of an art history book, and there you would find the image. The second method takes almost no bits, and it’s very fast.
Of course, the question this leaves open is, where is this shared dictionary? Where’s the dictionary that you can look up the movie in, but anyhow, Roel Pieper is in, and they’re all ready to turn this company into the next unicorn billion dollar venture.
But you know, revolutionary tech doesn’t just sell itself, you need money to turn prototypes into real products, and to pay people. Sloot’s team knew they needed major investment to commercialize his invention, and so they went knocking on some very high powered doors, doors that might have not opened before, but now would.
They met with the director of the Amsterdam headquarters of ABN Amro Bank, one of the biggest banks in the world. And they got a demo as well. ABN Amro bought 5 percent of the company. There were conditions, but they were buying it for 2.5 million. But Amro wanted collateral up front. Sloot had to deliver a patent application and source code to a lawyer to hold in escrow.
Without that, no money. And they needed to come forth with a major VC investor to lead the next rounds. It was time to put up or shut up. If Sloot Tech was for real, then putting down a patent would be a great way to prove it to the world.
I should mention, I got this information about ABN Amro from Eric Smit. He, in 2004, wrote the Dutch book, The Source Code. And when it started mentioning patents, I got excited. Patents are public information, and they’re generally pretty technical.
So it can help us get to the bottom of how this algorithm works. I never found the patent that ABN Amro demanded, and you’ll see why, but I did find Dutch patents, which were assigned to Jos Van Rossum, the first investor, and they described how the Sloot Digital Coding System worked.
So here we go. The answer to how the system is supposed to work is written down. It goes like this:
Each pixel of a frame in a video is turned into a number that represents the colors of the pixels. Nothing crazy there, right? Then the pixels for an entire line of video across the screen are combined together, right?
Multiply them or concatenate them into an even larger number, right? Then that number is stored. It’s converted into a key. It doesn’t say how, but I’m just picturing like a hash table. You put the value in and you get back out some hash that you can use to look it up. And then they just repeat that every line is turned into a number.
Every number is stored and given a key. And then the entire frame is made by combining all those and getting a key for that. And then every frame in the movie is stored the same way. And then presumably, that giant number of keys for every frame, the same thing’s done again, until you’re left with just one singular key that you can use to recursively look up all these levels in this hash table.
The whole movie just becomes a whole bunch of lookups. Now this would work. This is basically a data deduplication scheme, right? If a, if a line of video or an image in a video is repeated, this scheme would only store it once. You know, a, a totally black screen that might get cut to many times in the movie would only be stored in one spot.
But really, how many duplicate frames, how many duplicate lines are, are in a movie? It’s easy for me to see, though, how this idea came out of Repabase, his idea for storing PDFs, right? If every Philips manual has the same title page in this database, you would only store it once. You’re deduplicating all the pages.
But now we know what the key is, what’s stored on the smart card, according to Sloot’s patents, is exactly what he says it is. It’s not a compressed movie, but it’s a key. For doing this retrieval that you can use to reinflate into the movie. This kind of explains what’s going on. I think that the key for the movie is on the card and you know, it might be like 1305 and then it goes to some lookup table and looks up the movie that matches that.
And you know, the whole thing gets recreated. That’s what’s in the patent.
But that doesn’t explain nearly enough, right?
How did this work without the laptop? If he compresses a movie on his laptop and then hands over the card and plays it back on a device, how did that data get there? The patent does not contain the information about how the lookups work.
That’s what’s missing. The thing that he hinted at is that the lookup table was merely something solid state, merely a circuit. With no storage at all, a circuit that given a certain key could return the values, a magic solid state hash table that contained no data, which could turn any key into the value that was stored.
It’s as if instead of storing pi, he said he had this circuit that could return the offsets into pi at any point, instantly. So he didn’t have to store it. He could just pluck it out of the air. I mean, maybe he did. Maybe he did have that device.
But I still don’t think it could work. And the reason why, you’ll see in San Francisco.
Okay, so as the new CEO, Roel Pieper didn’t waste any time. Right? He left something big. He was gonna make this big. He wanted to strike while the iron was hot. And the big tech money was in California, it was in Sand Hill Road. So people lined up meetings with all the heavy hitters. Kleiner Perkins, Sequoia Capital, Sun Microsystems, etc.
This was their chance to cement major backing for their new company. They christened their company Fifth Force. I don’t know why, maybe they had this physics idea that they found a fifth force of nature that was information. Anyways, Pieper needed Sloot and his compression box to wow these investors, but Sloot didn’t want to go.
He claimed he was tired and he certainly was not going to let the device go without him. This was his baby, but Pieper wouldn’t take no for an answer. He told Sloot the Silicon Valley meetings were non negotiable for getting the major financing and they needed the financing, but Pieper wouldn’t take no for an answer.
He told Sloot the Silicon Valley meetings were non negotiable for getting the financing and they needed Sloot refused to budge. Pieper would have to reschedule, but Pieper offered a compromise. How about Sloot stays in the Netherlands, Pieper will just bring the compression device with him. But no, that’s not gonna work.
No one can be trusted with his life’s work, and certainly not this CEO with his big ambitions. The device had to stay with him, only he could operate it. Which strikes me as You know, a pretty odd assertion, a little bit of a red flag, but you can imagine how this went over with Pieper, the CEO who had left this role and was trying to dazzle these world class investors, but his secret weapon, his genius won’t make the trip, but he didn’t want to force them.
He didn’t want to endanger this brand new relationship by pushing too hard on this fragile genius. I mean, he’d seen it fall apart at Phillips. He wasn’t ready to push him that far, but he did need the tech demos. You know, he had those dollar signs in his eyes. He knew this was something. He didn’t want to kill the golden goose.
So eventually, he convinced Sloot that he could bring the device along, and he would personally supervise it. But in his paranoia, Sloot said he’d only agree if he painted the entire contraption with tamper proof paint. He wanted to be able to tell if somebody had betrayed him. Anyways, Roel jetted off to Sand Hill Road.
And the demos killed it. They were impressed. Rewinding through 16 movies on a monitor with no lag time, insane compression ratios. It was everything the investors wanted to hear. They were ready to write checks. But there was a catch, right? The big guys wanted to see the device proven out. They wanted due diligence and a controlled setup.
You know, encode a video in this one shielded room and then walk down the hall and then decode it in another shielded room, just using a chip card that we present and content that we present. We need a foolproof demo. But that would all have to wait for Sloot. Sloot was waiting anxiously back in the Netherlands, but things went well.
And when Roel came back and handed back the device, Sloot freaked out. He said one of the screws had been tampered with, that the VCs were probably trying to steal his idea. This was the ultimate betrayal, right? The tech vultures of America were trying to peek inside his box. But the thing I don’t get is, I don’t totally get what he was so afraid of.
In fact, going through all this data, trying to make sense of the story, one thing I often wondered is, is how things got this far. Because there’s flaws in, in Sloot’s scheme.
The Lookup Problem
Let’s revisit my Pi compression scheme, right? Suppose we’re exchanging data using this idea where instead of sending it, I send an offset into pi.
There’s practical problems with giving you an infinite amount of pi digits, but let’s just assume that I can. The issue is, the offsets become enormous. If I want to send you the offset for a 5 second clip of this podcast, I would likely find that that offset into pi, once I found it, which would be a challenge, if it even exists, would be larger than the data.
Like, sure, every hundred digit number might be in pi, but the offset for that number would often be much larger than the number it represented. In other words, since there are so many possible wave files out there, a key to look up a specific one would have to be very large. So even if Jan, as he claimed, had this chip where he could look up the constituent parts of movies into it, there’s still a problem.
The key to look up into that would have to be very large. I believe Jan thought this system could work, but I don’t think it can. And you’re about to see why.
Back in the Netherlands
Back in the Netherlands, Jan is tinkering away in his workshop in his attic, still trying to perfect some things with his device. He’s fueled by coffee and cigars and just mounting pressure to make everything work.
And I imagine him working there in like a yellow polyester shirt, you know, it’s sticking to his back because now we’re in July and it’s hot and it’s an attic and he’s fiddling with his prototype and he’s You know, examining the temper proof paint and getting angry. Getting more paranoid about how people are trying to figure out his device, but he doesn’t have time to obsess too much because the bank wants to wire the money, but he has to meet the requirements, right?
He needs to send the full source code and documentation for how this will work. You know, the beginnings of a patent application need to be put in escrow for ABN AMRO to pay out. So maybe this patent will explain how the system looks up keys. How this mapping works without an infinite key width. How all this can work without external storage with some sort of solid state circuit.
Maybe once and for all, he explains it into that patent application. But either way, Sloot is stressed and overwhelmed trying to put this all together, but he meets the conditions. I imagine Sloot’s lawyer placing the source code and the patent into a safe in escrow, and then Pieper finalizing the papers with the bank and the money being wired to the Fifth Force bank accounts.
But then, a couple days later, after proving himself to the bank people, Sloot gets an invitation from Roel Pieper. Come over to my house and come perform your compression demo. Bring the device. Be prepared. It’s a very important guest. Sloot drives his metallic grey Mercedes to Pieper’s lavish estate.
He’s nervous, obviously. Who is this person he’s supposed to impress? And inside, Pieper meets a smiling gentleman with owl like glasses, and it’s none other than Tom Perkins, the VC investor, the founder of Kleiner Perkins, one of the most influential investment firms in Silicon Valley. Roel and Perkins are tight.
Roel has been trying to sell Perkins on this tech for a while, but on the phone, Perkins just doesn’t buy it.
Roel, remind me, how much memory is on a smart card? It’s not much, is it?
Roel says 75 kilobytes. 75 kilobytes max.
Roel, that defies Shannon’s theorem. I think maybe Fourier’s theorem and Greene’s theorem too. It’s impossible.
Roel Pieper gives him the answer. He’s been giving everybody, It’s a key, it’s not compression, it’s a lookup. And probably I’m guessing the Kleiner Perkins people are excited about this invention, but they want to vet it. They want to get it in a controlled conditions.
And I’m just guessing here, but probably Roel says, you know what - my guy’s nervous, he’s not traveling to San Francisco. And so Tom Perkins does the only thing possible to put this tech through its paces. He travels to the Netherlands. I don’t want to miss out. But I need to vet this thing, so here I am. Let’s see it. Let’s see the validation.
The pressure is on, as Sloot wants to impress, you know, Mr. Perkins himself.
And Perkins has a demand. He doesn’t want just the standard demo. He wants him to tape a live TV segment and compress it onto a chip on the spot. An impromptu test of Sloot’s abilities. Sloot says fine, and he frantically starts getting the box ready to set up.
But his hands betray him. And he drops the device, a screw comes out, when he hooks it up, the image is flickering and he’s getting really stressed out and flustered. Perkins gives Sloot some time alone. Him and Pieper go off and do something else, and then they come back. Perkins has no doubt seen his share of nervous investors and he wants to give him a fair shake.
So eventually they come back, Sloot’s figured things out. He inserts a card into the computer, and transfers the video onto the card. And then, he transfers that card into the playback device, and there you go, a live recording being played back on the playback device. The monitor springs to life, and it plays back the show that Perkins demanded he record.
Perkins eyes widen. Rewind it, pause it, he gives him instructions, and Sloot does so. The demonstration ends, and Perkins grabs Sloot’s hands and says:
If this company works out, you’re gonna be the richest man in the world.
A few days after this demo, that went well, Sloot was found dead. He had a heart attack. His son found him in the morning, in the backyard in his house, outside where he smoked.
He probably had gone out at night to have a cigar, and he never made it back in. Now Sloot did have a heart condition, and he did often seem unwell. And there was a lot of pressure on him. But this whole thing is tragic.
Sloot was the key to everything. But the remaining team members, they told themselves, you know what? The company can survive. There’s the source code. There’s the patent application. They’re in escrow. There’s the demo devices. Everything seems fine. We can move forward. Our payday is coming.
But they opened the safety deposit box at the lawyer. And they only find a few papers describing the Repabase system, and some private documents, and then for some reason a book by Robert Ludlum.
And without Sloot understanding, this invention became a monumental task.
They brought in an expert, Marius Abel, a software guy Sloot had worked with in the past. And he analyzed the device, and then he searched every disk and every chip, and tried to find whatever he could. And he found nothing.
The Hard Drive
I mean, he found things, but he didn’t find how it worked, how this compression encryption scheme could work. And the more he studied it, the more something seemed off because the device that was left behind, it had a hard drive inside it. You could even hear it whirring when you used it.
But Sloot always said his compression didn’t need a hard drive. So that was strange. And then looking at the chip cards, it seemed like the data on the cards was just random noise. There wasn’t a scheme that Abel could understand. It was supposed to contain the key for the video, but it was just a mystery how that encoding worked.
But the big thing that was missing… was what Sloot called the compiler. The thing that produced the key that they put on the cards. That was missing. That’s what Abel couldn’t find. And investors started looking through everything that Sloot left behind. Notebooks, computers, file cabinets. If the compiler of this magic key generator was hidden somewhere, then they would find it.
But each new document was another dead end and they didn’t find the source code anywhere. Some on the team recalled that Sloot worked frantically on his laptop before demonstrations. Almost like he was compressing each video live, himself. But on the laptop there was no special software they could see, no compiler, or anything that could recreate his magic demo.
And so doubts crept in, right? Some at Philips had always thought it was a wireless trick, that things were being broadcast. But now there was a hard drive. And if Sloat had truly invented the impossible, how had he done it, and why had it now evaporated? If you read everything you can on this, and I feel like I have with the help of Google Translate, it seems like after Sloat’s death, the playback device still worked.
You could still play the 16 movies, but it also clearly had a hard drive in it. And that made it seem not like a magical device, but just some 1990s PVR. Some thought that this was a different device, an earlier device when hard drives were required. And the real one, the one he had demoed. That he promised didn’t have a hard drive, had been stolen.
Maybe this was industrial espionage that he was paranoid about. But no compiler program was ever found. And the supposed thieves didn’t take anything else. So it seemed very unlikely. And the company had to shut down.
What Really Happened
To me though, it’s clear what happened here. This was an inventor who thought he had a breakthrough. And he was still working out some of the details. He didn’t get it quite working and so he resorted to deceit, playing video off of a hard drive, or streaming it wirelessly, but probably just playing it off of a hard drive.
Jan’s paranoia was probably as much about not wanting others to discover the hard drive in the device and the fraud that was going on than anything else.
The thing, in fact, did what he initially said it did. The smart card contained a key, and the playback device would use that key to look up what movie to play from his Repabase database.
But the database isn’t magical. It didn’t encode all movies ever created. It just had what it had. That’s why most of his demos were just of these 16 movies stored on a card. He was able to play those back. And as far as I can understand, they could still play those back after his death.
How, then, did the live recordings work? How did he encode movies on demand? For Tom Perkins, I’m not sure. But maybe he faked a technical error and dropped his machine. He faked a flub so that he could stall for time and figure out a way to make the demo happen. I think it was misdirection.
I think Sloot thought his system could work, that he could find a way to store all possible movies in his RepaBase encoding system and look each up with a key, but he just hadn’t cracked it yet.
And yet, in the meantime, to get people interested, while he worked out the details, Yeah, he faked some demos. He said it’s not compression, this is a key to unlock the movie. But he missed the fact that that could never work with arbitrary movies. If that system stored all movies, all possible frames, all possible pixels, Instead of just 16, the key would need to be much larger.
Or to put it another way, there are more possible movies than can be represented with a one kilobyte key. A one kilobyte key is like less than one bit per frame of a two hour movie. It’s like trying to give a unique name to a billion people using just a thousand names. No matter how you rearrange it, you can’t avoid some people getting the same name.
There’s just not enough cardinality. There’s just not enough variability in those keys to represent all the data that’s needed. His system was mathematically impossible, and he just didn’t know it.
The Quest For the Holy Grail
So that’s my take on the Sloot system. Sloot faked the demos because he didn’t get the mathematical impossibility of his claims. And despite warning signs, investors were swayed by his persuasive demos. Really, it’s a cautionary tale of greed overriding skepticism.
I talked to Eric Smit, the investigative journalist, who is the source of most of the data for this episode, and he had a slightly different take.
It’s very believable, possible, that he could have very much exaggerated his invention, but nevertheless, he was able to show something that was very, very convincing, and made people, people who were into this technology and knew about it, made them believe.
My opinion is, is, is that he had at least something that was jaw dropping and that is, that, that’s the only thing I’m fairly certain of, yeah, listening to all the people I spoke to.
And also, Eric says, it wasn’t just Sloat who experienced bad luck. After Jan’s death, Roel Pieper and his wife were attacked. A crazy man stabbed them and they ended up in intensive care. They survived, but just barely. And then around the same time, Tom Perkins car exploded and he barely survived. All these things going wrong, it led to conspiracy theories about this device.
To Eric, this is a common thread in these types of stories:
So I think, I think of the story also as a typical grail story. Everybody’s running behind something that has incredible potential. Like if you can take a drink from it, you get eternal life. In this case, it would be eternal richness, uh, richness. Roel Pieper started understood straight away. He saw that invention as his search for the Holy Grail.
He says, well, if I can, I get ahold of this and I can become CEO of this company. And in the end, uh, everybody dies or, you know, things don’t end up and well, especially not…
Especially not for the guy who invented it.
Adam: That is our show. I hope you enjoyed the story. I hope you learned a bit about compression.
Big thanks to Eric Smit, the investigative journalist who broke this story back in 2004 in his book and even earlier in magazines. So much of the detail I have here comes from his book. It’s by far the best source of any of this information, unfortunately, only available in Dutch. But anything wrong in this episode, that’s on me. Anything. You know, correct. That’s the great reporting of Eric. He wanted to be more involved in this episode, but a family emergency got in his way.
If you speak Dutch or there’s even a German translation, you should buy his book. He’s now the co editor and chief of follow the money, a Dutch investigative journalism website.
Also, if you like the podcast, you might like the newsletter. I cover similar topics there or follow me on Twitter @AdamGordonBell. I often share behind the scenes details of the podcast or just other things going on.
And for true fans, the best thing you can do to support me and the CoRecursive podcast is become a supporter. There’s a link in the show notes and join as a podcast supporter, you’ll receive bonus episodes and you can join our community.
I’m hoping to get Eric on for a bonus episode. He’s got details about Sleuth that didn’t make it in the book. Stuff we still don’t know. And he’s looking through his files.
So if you’ve got questions for Eric, send them my way and we’ll get those answered.
And until next time, thank you so much for listening.