Expensive Dreams
I woke up this morning to a 100% loss on one of my positions in the stock market amounting to about 3% of my net worth. I don’t think that’s a terribly unusual experience for an active trader like me but it’s certainly jarring. I thought I’d try the “writing therapy” thing and talk a bit about it.
As we approach the end of the current credit cycle there’s likely to be a lot more of this. Typically towards the end there’s a lot of FOMO both from more careful market participants who are under performing and from people who have no idea what they’re looking at and turn to gambling as a last resort. And of course at the very end the rot is shown for what it is.
Here’s my attempt to rationalize why I and the rest of the world seem to disagree so severely about this.
America and the Dream Industry
Dreams are the process of looking beyond the frontier and imagining the possibilities that come from being a pioneer. Being a pioneer is the process with which these dreams are realized. Here there is rot, decay, want. There we could have abundance and understanding. The dreamer finds a crossing and the pioneer builds a bridge from here to there. The dreams must come first, then the exploration.
Computing was a dream.
America was a dream.
The Web was a dream.
LLMs were a dream.
...
The list goes on. In many ways America’s primary export is dreams. Our largest industries: media and technology, serve the purpose of encouraging people to dream and then monetize the realization of dreams. America was not only founded on Dreams but runs on the dream utility.
Our most well known PE funds like Ycombinator exist entirely to facilitate dreamers realizing their dreams and their business models are built on the expectation that most will fail.
Most Dreams Fail
The dreamer is a prophet and oracle. The act of looking beyond the frontier should be impossible by definition. Dreams fail so often there’s also an entire industry which capturing value from the destruction of dreamers.
It’s well known among biotech investors that something like 10% of drugs actually make it through clinical trials. Most drugs simply don’t work at all and probably the fastest way to make a lot of money is by reading the studies yourself and finding biotech companies to short. N-gate ran for years mocking HackerNews (I was under the impression the original owner actuall managed to sell it though I’m not sure.) Coffezilla, and many other YouTubers call these things out from the sidelines. Entire hedge funds run on short selling.
Dreaming May Be a Skill
Looking out over the horizon like this isn’t an entirely gnostic practice. Careful understanding of particular reality enables people to guess at what might be unseen.
The ancient Pacific Islander culture that produced balanced rig sailing did exactly this by looking for interference patterns in ocean waves to find far away islands.
Steve Wozniak dreamt of the Apple computer while spending his days programming minicomputers made of similar components.
The researchers at Bell labs dreamt of the transistor as consequence of their understanding of material physics
The Philosopher’s Stone
Let’s talk about the transistor for a moment. It’s the end of millennia of scientific research. It’s the realization of the dream of the ancient alchemists. Pure crystals with patterns etched into them that can communicate with eachother using RF energies, can hold ideas, can animate automations. These are the philosopher’s stones the alchemists dreamed of.
Alchemical culture did not realize the dream of the transistor however. Alchemists were not good dreamers. They engaged in wild fantasies and stabbed at the frontier blindly. The intellectual descendants of Lord Bacon, Micheal Faraday, and Friedrich Wohler did after a great deal of care and observation. While from the outside dreaming appears to be the beginning of a journey in a real sense it’s the end of an older one of observation.
The nature of knowledge work
It’s all in your head. Just like physical labor you can take shortcuts, you can skip important preparatory work, you can make it look like you’ve done everything when you haven’t. Unlike physical labor though since the artifacts dreamers produce are dreams it can be horribly difficult for someone who does not share the dream to understand when they’re speaking to someone genuine and when they speak to a charlatan.
In the modern world we’re surrounded by software and finance. These linguistic artifacts do things normal people could have never imagined as a consequence of someone’s technical understanding of what’s possible.
Look how programmers, people who’s purpose is to understand networked computing, collaborate. Typical computer users will want to collaborate on a document by sharing updated versions via some communication system, or if they’re more “hip” might use a “multiplayer” CRDT or OT system like Google Docs or Figma. In their mind nebulous “technology” will synchronize the changes for them.
Programmers do things other people cannot dream of because we understand computing. Here programmers dream, they understand the limitations of synchronizing changes across an unreliable and potentially partitioned network so they know the workflow has to change. Instead a programmer uses a “branch/merge” workflow and DVCSS tools to support it.
The Trolls Who Consume Thoughtless Dreamers
All of this allows for a kind of confidence trick. In the old days a fraudster might write a purchase agreement for some commodity he cannot possibly acquire and then abscond with their counterparty’s money when he fails to deliver.
In this way our conman dwells under the bridge from here to there and sells a false dream to careless dreamers.
History provides us numerous examples of these people.
Thai “here in my garage” Lopez famous for buying ads where he literally monologues about dreams
Enron who told us we could offload energy volatility to speculators and then fraudulently performed market making operations in their own order book.
FTX who did the same with crypto currencies
pets.com and friends who sold investors on the idea that together we could all make money with just an online storefront and a mediocre business.
My Struggle with D-Wave
So on the bridge from here to there, being a dreamer myself, I saw the troll and called it evil. I went to Github and drew the sword of truth. They claim to have ML so I read their pytorch code [1] . All that’s there is a boltzmann machine optimizer, it cannot generalize. I programmed against their optimization code and read their papers, I was even up until 1:00 AM last night writing a short program to compare simulated annealing which they keep holding up as the apex of classical optimization to a modern gradient based algorithm. it became extremely clear to me that they actually had nothing to sell and could not ever have anything to sell.
There was actually an attempt by the Chinese to play a very similar trick just the other year. They repeatedly claimed to have achieved functional quantum computing that outperformed classical computing. Closer inspection revealed however that the only problem their machine could solve (boson sampling) was simulating itself; They were correct in a tautological sense but the practical reality was that they had no product. The Chinese tend to not dream so big, they’re much more concerned with actually doing, so this scam was very short lived.
Here in the US though, not so. Listen to any d-wave shareholder, if you can even get them to argue their position, they’ll tell you about the most fantastic ideas. The last one I spoke to was talking about doing mellenox over quantum networking. Now I’ve actually been paid to write code for mellenox cards. I can tell you right now this is complete gibberish but they were literally, physically breathless with excitement. The primary advantage of quantum networking is a secure but slow media layer (anyone who knows anything about networking is giggling at that) and the primary advantage to mellenox is a highspeed DMA capable bus for peripheral cards. The idea of having any security beyond maybe IOMMU for mellenox is absurd. The idea that there’s value add of any sort here is impossible to find.
This man is a thoughtless dreamer who has been sold a dream by a confidence trickster. He had no idea what he bought but was sure it was the future. This man has been eaten by the troll that lives under the bridge from here to there.
And D-Wave is one of the worst, their machines have no hope of even these things. Like the Chinese machines all they can solve is the Ising spin glass problem (which is mathematically identically to QUBO.) They have no real world application
So what could I do here? I’m unemployed, scientifically literate (so I like to think) and armed with vim and a margin account I hopped under the bridge and threw myself into combat. I sold at 10, 12, 17.
Two quarters went by with nothing to show but work on a fiction that could never materialize, Last earnings call they mentioned VAE code out of nowhere. I flew to github knowing their machine wasn’t even mathematically capable of running anything like that. Sure enough a month old PR added encoders and decoders running on classical computers and they had jury rigged a latent Boltzmann machine in the middle. More technical truths that were effectively lies.
18, 20, 25, and now people are trading D-Wave at 32. After yet another paper where they embedded a QUBO problem into the machine, compared a “hybrid” solution to simulated annealing (a very old and slow optimization algorithm) on a classical computer, and claimed to have outperformed classical computing.
Looking at the order book there are still more buyers than sellers. Eleven billion dollars given to charlatans who lie to people because they know their shareholders can’t understand what they say. I’ve been overwhelmed by the weight of the absolute horde of retards flinging themselves into poverty, I had to relent.
Here I am, the intellectual who’s lost his argument. The troll got me too.
And That’s a Good Thing
As costly and frustrating as the dreams of an idiot are, what does the world look like when people lose the capacity to dream? The zillenials and zoomers know. when people rot in bed with no reason to leave they do that because they’ve closed their mind’s eye and see nothing but the particulars of the frontier closing in on them. They say “There is nothing in the world for me and so I have no reason to get up and do things.”
I think this is what frightens me so much about this sort of thing. The conmen and enemies of capitalism are the enemies of our capacity for corporate dreaming itself. I think also that’s why Americans (and surprisingly enough the American proletariat) reacts so negatively to communism. They understand what’s at stake here.
This is why gamblers are confusing, we don’t understand what he is buying. The gambler isn’t trying to buy future money and expecting to be paid a term or risk premium. He is buying a dream because he’s short of dreams himself. A shortage of dreams is a kind of living death worse than any material suffering. This is why hordes of immigrants from the third world assault us. This is why boomers throw everything away in vegas, this is why unimaginative (or perhaps overly imaginative and hungry for dreams) people throw everything away in the stock market.
In the same way value investors think of equity as a derivative of revenue rather than of this corporate dreaming process. That’s why we’re so often wrong. Debt is the revenue derivative.
Fundamentally I think my mistake here, like the buyers, was misunderstanding what I sold. I didn’t sell a purchase agreement on fictional compute, I sold a dream. I sold a dream that couldn’t possibly exist in the real world and so I could never hope to deliver in the future. As long as they can keep the lights on and telling shareholders they’re working on the dream machine people will keep buying the stock, and thanks to the extraordinary work of dreamers before us lighting is not very expensive.