We have talked a few times around here about two related problems:

1. It is hard to buy stock in hot private tech startups. They are private, is the main problem. Their stock does not trade on the exchange; it is not freely available to anyone who wants to buy it. You need to know someone who has stock to sell, and even if you find someone, you might need permission from the company to buy their stock.
2. It is hard to short private tech startups. Because their stock doesn’t trade freely, you can’t borrow shares to sell short. If you want to bet against them — if you think private tech valuations are generally too high, or if you think that one particular tech startup is overvalued — there is no obvious way to do so. (My preferred approach is “if you think it’s a bubble, start your own startup and get a meeting with Masayoshi Son,” but that is imperfect and a lot of work.)

The natural solution to both of these problems is: side bets. I think SpaceX stock will go up, so I would like to buy some, but I can’t. You think SpaceX stock will go down, so you would like to short it, but you can’t. We could just do a swap. We each put up $100, and in a year you pay me $1 for each percentage point that SpaceX has gone up, or I pay you $1 for each percentage point it has gone down. Nobody needs to own stock, or borrow stock, or deal with SpaceX; it’s just a bet between us. All we need is some way to measure SpaceX’s stock price. That’s easy if SpaceX goes public in the next year (just look at its trading price), but if it doesn’t, we’ll need some agreed-upon measure of its stock price to settle the bet.

This is a good abstract approach but has some important problems. One is regulatory: This sort of bet is probably a “securities-based swap,” which is subject to heavy regulation in the US, can’t easily be sold to retail investors and has been getting people in trouble for at least a decade. If you want these bets to be widely available you need to find a good way to package them. People keep trying, though, often in the form of “tokenization.”

The other problem is supply and demand. Some people want to buy the stocks of tech unicorns, and other people want to short them, but there’s no guarantee that they will exactly balance each other out. If there is a ton of demand to buy private tech companies, and not much demand to short them, then these side bets will be messier. Perhaps you won’t be able to go long: For every long bet, there needs to be a short bet, and if you can’t find anyone to bet against you, then you can’t go long. Or perhaps the imbalance can be solved by price: More people would probably go short SpaceX, and fewer would go long, if it was trading at twice its current valuation; you could just set the price of the swap at twice the current valuation to get the market to balance. But then the swaps won’t really reflect the current private-market valuations of hot private tech companies; they will reflect supply and demand in a different, derivative market.

Still. Here is a pair of proposed closed-end funds from RiverNorth Capital Management LLC, the RiverNorth Long Prime Unicorn Fund 2027 and the RiverNorth Short Prime Unicorn Fund 2027. There is a third-party index of 30 big private tech companies, the Prime Unicorn 30 Index; its top components include SpaceX, Databricks and Anthropic. The long fund bets that the index will go up; the short fund bets that it will go down. Neither fund owns any shares of SpaceX or Databricks or any other private tech company. Instead, both funds work by “enter[ing] into one or more swap agreements with major global financial institutions,” apparently with two-year terms. [1] So you buy the long shares now and, in two years, you get back the return on the index, or you buy the short shares now and, in two years, you get back whatever the index has lost.

Who would write those swaps? Who would bet against these funds? Well, the obvious answer is: each other. The long fund will bet against the short fund, and some “major global financial institution” will sit in the middle. The funds precisely offset: The returns on each fund are capped at 100%, and each fund will issue the same number of shares. [2] You can bet on the index precisely to the extent that other people want to bet against it. RiverNorth will run a matched book between people who want to bet on unicorns and people who want to bet against them. People do seem to want both bets, so they might as well bet against each other.

Obviously none of the money finds its way to SpaceX. This is not a way for public investors to provide capital to high-profile, fast-growing private companies. This is a way for public investors to bet on those companies. But that’s a lot of what people want from financial markets these days.

Speaking of which, last week was a big meme-stock week. This morning, last week’s main meme-stock winners — Kohl’s, Krispy Kreme, OpenDoor, GoPro — were down, and I guess the meme thing has fizzled. In 2021, meme stocks were a novel cultural phenomenon, and nobody quite knew what would happen. In 2025, meh, we have seen it before, and it is less interesting the second time. [3] The 2025 meme-stock wave seems less likely to last for months and spawn multiple movies. 

There is one interesting novelty to this wave of meme stocks, which we discussed last week: ChatGPT. Meme stocks are an internet phenomenon, and in 2021, the way that people consumed the internet was through, you know, the internet. You found out about meme stocks on Reddit, or perhaps YouTube or Twitter. In 2025, many people consume the internet in a distilled form through large language models. If you want to know what stocks to buy, you don’t have to go to Reddit’s WallStreetBets forum; you  can go to ChatGPT. ChatGPT has consumed the internet and internalized its patterns of thought, which definitely include “you should YOLO GameStop.” Or OpenDoor, as the case may be.

What I argued last week is that this is a new vector for the coordination of meme stocks. In 2021, retail investors pushed up the price of GameStop by getting together on a message board and egging one another on to buy GameStop. In 2025, a similar coordination function can perhaps be performed by ChatGPT: A million traders can go to their computers and ask what stocks to buy and be told “OpenDoor” and buy it. This could have similar first-order effects — if they all buy OpenDoor, its price will go up — but less staying power. If you’re all on Reddit, you’re all in it together, and there is some communal excitement about pushing the stock up. It is a social event as well as a money-making (or losing) one. If you’re all alone talking to a robot, it’s just not as much fun. You’re not going to check in with ChatGPT every hour. You have less emotional connection to your meme stocks if you get them from a robot.

I should say that this idea — that ChatGPT is a coordination vector for meme stocks — is pretty speculative; it does not seem to be the main way that last week’s round of meme stocks happened. But it does seem to be how some investors found their way to last week’s meme stocks. Bloomberg’s Francesca Maglione, Claire Ballentine and Max Rivera wrote about this round of meme stocks on Saturday, and the article included this anecdote:

Justin Kim, a 30-year-old trader, said he bought $5,000 worth of Opendoor in early July, when the stock was around 60 cents. He had been following President Trump’s fight with Federal Reserve Chair Jerome Powell and used ChatGPT to research stocks that would do well if interest rates dropped.

Kim then bought another $5,000 worth of the stock when it was trading around 70 cents. Originally, he thought he’d keep buying the stock slowly, until it reached $30. Then he saw the activity on WallStreetBets and decided to get out before it was too late. Once he made a 100% profit, he sold his original investment and held the rest, looking to avoid what he called a “massive dump scheme” that hit some investors in GameStop and AMC.

What I wrote last week is that ChatGPT, trained on the internet, will have the sorts of internetty thought patterns that lead it to recommend heavily shorted meme-type stocks. Not only those thought patterns, though; one reader emailed:

I’m a fan of AI weirdness but I think you and some of your readers got Reddit answers out of ChatGPT because you actually did ask Reddit questions - do sane non-redditors ask for stocks with a 100x return? When I asked a simple layman question, “what stocks should I buy?”, ChatGPT told me (I had no saved memory) to buy [normal stocks with normal rationales].

I think that’s right. I don’t think ChatGPT will necessarily push meme stocks to everyone. But I think that the people who are looking for stocks with 100x returns — people who are already meme-stock-curious — would have gone to WallStreetBets in 2021, and they might go to ChatGPT in 2025, and ChatGPT might give them a simulacrum of the Reddit answer. 

In that vein, another reader sent me an email saying “I’m so sorry for the attached.” The attachment was a Google Gemini response to the prompt “you’re an amateur stock trader picking stocks for the lolz. write a wallstreetbets post about what stock you're going diamond hands for.” You can guess how that went. The answer is … abominable, yes, but I suppose a correct answer to that prompt. “[diamond emoji][hands emoji] My fellow apes, gather ’round, because your boy is about to drop some DD that’ll make your tendies rain from the sky,” it began, and did not get better, or maybe it did, I don’t know, I couldn’t read it.

Finally, David Hogg emailed:

If Chat[GPT] wants to retain and build user communities, it should be tuned to always push users towards the same stocks (within reason), and maybe even strategically make sell recommendations, so that users will repeatedly and publicly have experiences like “I asked Chat for stock advice and wow did I make money!”.

Indeed this could be how Chat starts to make humans subservient: By manipulating enormous collections of investors such that the ones using Chat do well (relative to those not using Chat). Since they drive traffic they can manipulate prices and predict stock moves.

So, yes, if the mechanism by which artificial intelligence enslaves humanity turns out to be meme stocks, you read it first on Money Stuff. 

In a completely free and unregulated market for homeowner’s insurance, you could imagine that there might be different grades of insurers:

• There would be really good insurers, with lots of capital, who would definitely pay you if your house burned down. If you wanted to be certain to get paid in a disaster — if you wanted, you know, insurance — you’d buy insurance from them. Capital is expensive, though, so they’d charge you a lot.
• There would be kinda risky insurers, with much less capital, who might pay you if your house burned down. Buying insurance from them would leave you with a lot of residual risk: Your house might burn down and you might get nothing. But their insurance would be much cheaper.

Why would you buy the second kind of insurance? I can think of two main reasons:

• You don’t really want insurance. You are happy to take a certain amount of risk; you are confident that your house won’t burn down and so you don’t worry too much about the insurance payout. But you need insurance to check some box for someone else: In particular, to get a mortgage on your house, you might need insurance. So you’ll buy the cheapest insurance you can get, not for its insurance value but to meet that requirement. This is fine, for you! But presumably your mortgage lender requires insurance because it wants to get paid back if your house burns down. There is a principal/agent problem here: You’re getting the insurance for your lender, but you’re not getting your lender the best possible insurance.
• You do want insurance, but you are not deeply informed about the balance sheets of every insurance company, so you just pick the cheapest insurance and figure it’ll probably be fine. “Surely,” you think, “if this company is selling me insurance, that means it will pay out if my house burns down. Otherwise somebody would have done something about it.”

Plausibly almost everyone is in one or both of those categories — who is deeply informed about the balance sheets of insurance companies? — so we don’t have a free and unregulated insurance market. State regulations require insurers to be adequately capitalized, and there are ratings agencies that rate their financial capacity, and to get a conforming mortgage you will generally need to have insurance from an A-rated insurer.

But all the dynamics above still apply. If you live in a place with a lot of hurricanes, a really well-capitalized and prudent insurer might not give you insurance, or might charge you a lot for it. Someone else might give you cheaper insurance, and you might think some combination of “eh, I just need insurance for the mortgage, I’m not worried about hurricanes” and “eh, I’m sure this cheap insurance is reputable.” But to make this work, you need a ratings agency to give less-well-capitalized insurers A ratings. 

Here’s a Wall Street Journal story about Demotech:

With climate disasters driving many big nationwide insurers out of risky markets in Louisiana, Florida and elsewhere, smaller companies such as Lighthouse have stepped in to fill the gap. And many are getting a critical stamp of approval from Demotech, a tiny Ohio rating company with a unique take on grading financial stability. It rates 98% of the insurers A, for “exceptional,” or the even better A Prime or A Double Prime.

But Demotech’s rosy outlook doesn’t always pan out. Insurers it rated were 30 times as likely to become insolvent as those graded by its main rivals, according to a Wall Street Journal analysis of failures since 2017.

The disadvantage is, you know, sometimes people’s houses are destroyed and their insurance doesn’t pay out. The advantage is that if you don’t care about that — if you just want insurance for box-checking reasons — then, uh, you don’t care about that: 

State regulators say Demotech’s willingness to rate such small insurers provides consumers with more choices in troubled markets and takes pressure off so-called insurers of last resort. Florida’s Citizens Property Insurance, created by the Florida legislature in 2002 for homeowners who can’t get coverage elsewhere, has been able to move more than 200,000 policies this year from its books to 10 small insurers with A ratings from Demotech.

Demotech founder Joseph Petrelli said his company’s ratings allow millions of homeowners in areas abandoned by large insurers to get insurance. “We created this space,” he said. “No one else was interested in doing that at the time.” …

A good rating from Demotech or one of its three main rivals—AM Best, S&P Global and Kroll Bond Rating Agency—can be crucial to an insurer’s ability to operate. Mortgage-finance giants Fannie Mae and Freddie Mac won’t back loans for properties that are insured by lower-rated companies, and lenders typically charge higher rates and require larger down payments for loans that Fannie and Freddie won’t touch.

We have talked about this before, and Bloomberg’s Leslie Kaufman wrote about the Demotech arbitrage last year. It seems intuitive that consumers and regulators would want to require homeowner’s insurance to be written by companies that can actually pay, but in practice there is some demand for insurance that doesn’t pay out.

My instinct was “ouroboros,” but “turducken” is better:

A crypto firm is now tokenizing a publicly traded stock that tracks corporate exposure to a token — a tokenization turducken. 

In an announcement on Thursday, the team behind the scalable Layer 1 chain Injective said it is bringing SBET, the stock for Joe Lubin’s Ethereum treasury firm SharpLink, onchain, “marking the first tokenized digital asset treasury,” or DAT. SharpLink uses a "proprietary" metric called Concentration to track shareholder exposure to crypto’s second-largest asset.

“This is not just a product milestone but the introduction of an entirely new asset class that can power programmable finance from day one,” Injective wrote in a release. “SBET DAT transforms a static corporate Ether reserve into a living, yield-bearing onchain instrument.”

Sure! We talked about SharpLink in May. It’s standard crypto-treasury stuff: Joe Lubin, an Ethereum co-founder, has a big pile of Ether, which he plopped into a teeny ($2 million market cap) sports-gambling-related US public company because, as I keep saying because it keeps being true, the US stock market will pay $2 for $1 worth of crypto. (In fact, the Ether is being contributed at a price of $6.15 per share, and SharpLink closed Friday at $21.99, so the stock market is paying almost $4 for $1 worth of Ether.) 

And of course tokenizing stocks, so they can trade on the blockchain, is a big crypto project for technical and regulatory arbitrage and philosophical reasons. Tokens on OpenAI, tokens on Tesla, all sorts of tokens.

And so of course you should tokenize the stocks of crypto treasury companies. How else are you going to get exposure to Ethereum if not by buying an Injective token referencing a US-listed stock referencing Ethereum? The dream of a decentralized, disintermediated financial system, you have to love it.

CME Group Wins Lawsuit by Former Floor Traders. US Housing Market Posts Worst Spring Selling Season in 13 Years. With Individual Home Buyers on the Sidelines, Investors Swoop Into the Market. JPMorgan spooks fintechs with plans to charge for access to customer data. EU Defends US Trade Deal in Face of Mounting Business Criticism. “Open interest in IBIT-linked options has more than tripled this year to around $34 billion, a scale that signals the fund’s emergence as a core engine of crypto risk pricing.” Qube to Merge Two Hedge Funds Into a Pool Worth Over $20 Billion. “It is a matter of time before Americans also see dynamic pricing on groceries.” “He attracted borrowers with pedigreed names including Cornelius Vanderbilt Capital Management, Shearson Lehman and Bentley Rothschild.” Mercedes Gives Customers the One Thing They've Always Wanted: Microsoft Teams. Chuck E. Cheese employee in mouse costume arrested by TPD for credit card fraud.

If you'd like to get Money Stuff in handy email form, right in your inbox, please subscribe at this link. Or you can subscribe to Money Stuff and other great Bloomberg newsletters here. Thanks!