Say goodbye to the Previous Thing. Also: Salesforce gets f*cked by AI, Tesla’s Roadster scam, and more.
Welcome to your weekly Dark Markets roundup of technology investment fraud. I’m David Z. Morris, longtime financial journalist and author of Stealing the Future: Sam Bankman-Fried, Elite Fraud, and the Cult of Techno-Utopia.
In Today’s Edition:
* Scholars explain why LLMs are Built to Destroy Institutions
* Tesla’s biggest scam is … The Roadster?
* Meta shuts down the Metaverse
* Salesforce’s AI Pivot is a Disaster
* The self-destructing spiral of Prediction Markets
Subscribe
The ever-vindicated Gary Marcus runs down a new study from Boston University, “How AI Destroys Institutions,” that unpacks why “AI’s current core functionality — that is, if it is used according to its design — will progressively exact a toll upon the institutions that support modern democratic life.”
Specifically, they highlight that “The real superpower of insttutions is their ability to evolve and adapt within a hierarchy of authority and a framework for roles and rules [that] empower individuals to take intellectual risks and challenge the status quo.”
David Z. Morris
·
July 1, 2025
Read full story
Generative AI unfortunately does the opposite of that, simply regurgitating existing midwit understandings of the world, without any possibility for progress or innovation. One acute (if complex) example is the collapse over the past two years of Stack Overflow, once a community of people who shared knowledge peer to peer. The platform has been killed by AI, after the machines strip-mined the human knowledge already accumulated there. A common response among coders has been “Stack Overflow sucked anyway,” but that might just be one instance of the fact that “friction” is often a synonym for “community.”
Knowledge only ever advances through the meeting of minds. AI is putting an end to that.
Meanwhile, a new poll by Civiqs (an agency I’ve never heard of) claimed to find that 54% of American respondents had a negative view of AI, and only 17% had a positive view.
David Z. Morris
·
May 5, 2024
Read full story
I keep becoming a bigger fan of Wall Street Millenial, and their recent video on the Tesla Roadster was an absolute banger. I had nearly forgotten that Tesla started taking preorder deposits for a promised new version of the Roadster sports model back in fucking 2017 and still hasn’t shown a production-ready prototype.
Elon Musk has mostly stopped talking about it, except when last year he repeated the absurd claim, first made in 2019, that the new Roadster would literally be rocket-propelled. If you ever forget this guy is just a slightly more talented habitual liar than Sam Bankman-Fried, remember this howler.
Those included a number of $250,000 deposits for the “Founders Series” (pro tip: If you have to hand over a quarter mil, you are NOT a founder!), and WSM claims all the deposits totaled around $300 million. Moreover, they make the very compelling argument that the announcement and preorder came at a moment of acute financial pressure for Tesla, strongly suggesting that this was planned as a pure, shameless cash-grab from the very beginning.
Why aren’t the victims a lot more angry here? That’s a question that could be asked about nearly everything Elon Musk has ever touched.
Meanwhile, the FT dives into the criminal networks running massive catfishing and investment fraud operations out of Cambodia.
Meta has fired about 10% of its Reality Labs staff, shut down several VR studios, and entirely closed Horizon Workrooms, its Metaverse for Work.
Remember that the entire “Metaverse” pivot back in October of 2021, including renaming the company from Facebook, was not driven by any overwhelmingly convincing business case. While notionally there was some sense that “Gen Z likes Fortnite so let’s make a social network that looks like Fortnite,” the real goal was to distract from and minimize the Facebook brand’s tarnishing by election manipulation scandals in 2016 and 2020.
I’ve also maintained that the term “Metaverse” was lifted, like the prior idea for Facebook’s Libra token, from cryptocurrency. It wasn’t widely adopted even in crypto, but the term “Metaverse” for a time referred to the concept of an interoperable system that let you use blockchain assets – particularly NFTs – across different games or virtual environments. For instance, a “sword” NFT would have different attributes and abilities in different settings. That’s why it was “meta” – it implied a system of systems, an interlocking but ultimately open ecosystem.
That of course has nothing to do with what Facebook tried to build: a completely closed garden, mostly aimed at creating even the most superficial reason for someone to buy a VR headset. Most amazing of all, it was unclear to the end what you were supposed to do in the Metaverse: kids liked Fortnite, after all, because it was a game that let you shoot your friends in the face. No such killer feature existed in Horizon Worlds.
Subscribe
The reason Mark Zuckerberg was free to spend billions of dollars distracting from the failures of his core business is that he has extraordinary levels of control over his own company. With Llama AI shaping up to be another case of Zuckerberg trying to do the New Thing and failing badly, Meta is becoming an object lesson in the failings of this sort of corporate autocracy, which also describes controlled and powerless boards like Tesla’s.
Individuals unanswerable to even the most superficial sanity checks tend to make worse decisions than those working collaboratively, and when you’re talking about a $1.5 trillion company – or a country – the blast radius of these bad decisions is huge.
This is also happening in AI, obviously, with a few loud, persuasive figureheads leading us all down the drain. One recent update on the ongoing failure of the LLM-megascaling project comes from Will Lockett, who unpacks the specific idiocy of Salesforce, another firm with a charismatic and singularly powerful CEO. That CEO, Mark Benioff, “indiscriminately fired 4,000” staffers, or nearly 50% of Salesforce workers. It has predictably been a disaster.
Among other problems Lockett describes based on internal sources, “[Salesforce] employees have to spend so much time stepping in to correct the wildly wrong AI-generated [outputs] that AI is wasting more time than it saves. In other words, this AI reduces productivity, not increases it.” Having gotten rid of thousands of people who know what they’re doing obviously isn’t making the situation better.
If we keep seeing these reports of the AI headfake leading to massive disruptions and waste, well, at a certain point we’re going to be talking about the entire economy.
The beloved FT Alphaville is on Substack now, and this week they have an amazing breakdown of why prediction markets are a dead-end business model. Noting recent news that large hedge funds are entering prediction markets, they predict that the arrival of professional arbitrage and actual insiders will go the same way it did in the online poker space.
David Z. Morris
·
October 25, 2023
Read full story
Those poker platforms (including the tangentially FTX-lined UltimateBet) were fun places for aspirational players for a short period, before professional sharks and whales arrived to “farm” those less savvy “fish.” Eventually, there were more whales than fish, and the fish got tired of being sucked dry by more experienced players – an implication of what’s known as “adverse selection.”
In short, even without the cheating scandals and the eventual regulatory crackdown on the poker sites, they were doomed. Alphaville makes the intereting point that prediction markets are even more likely to be extractive from fish than meme stocks or crypto, because prediction market outcomes can be confidently predicted by experts with real information about fundamentals. That’s also basically the case with poker.
Dark Markets is a reader-supported publication. To access every post and support my work, consider becoming a paid subscriber.
