Thursday links: A battle of wits... and am I Satoshi?

The Wall Street-ification of prediction markets

Kalshi CEO Tarek Mansour argues that prediction markets offer a more level playing field for participants than traditional markets. As evidence, he points to the payment for order flow (PFOF) in equity options: “Citadel is paying your broker PFOF because they know you’re doing something dumb.” 

This feels a little rich, seeing as market makers have been active on prediction markets for some time now: Kalshi itself announced the onboarding of Susquehanna as a liquidity provider in 2024. 

(From personal experience, I can attest that Susquehanna is at least as adept as Citadel at identifying traders who are about to do something dumb and then helping them do it.)

Still, Mansour says retail traders are more likely to have an edge in prediction markets than traditional markets. He notes research from the Fed finds prediction markets are more accurate than economists in forecasting things like CPI (with the implication that retail participants are making the predictions).

He also cites a survey in which prediction market traders say they avoid the stock market “because it’s rigged against me.”

In short, he frames prediction markets as “less gambling” than equities.

I’d counter that prediction markets are negative-sum (the definition of gambling) and equity markets are positive-sum (the definition of investing). 

Either way, though, prediction markets are starting to look more like both Wall Street and Las Vegas. 

Beyond market makers, we also have prediction market “notes” — the first instance of which is a bond-like security that pays a 7% yield to holders if Nvidia remains the world’s largest company for one year. The issuer of the note hedges its exposure in prediction markets in a way that earns more in profits than the note pays in interest.

Pretty clever! Also, very Wall Street.

We’re also starting to see leverage in prediction markets. This will be restricted to markets that are still some time from resolving, so I’m not sure how popular that will prove to be. But I can’t think of anything more “gambling” than making a binary bet with leverage.

(Actually, I can: lending to a binary bet with leverage.)

Finally, new research proposes a way for prediction markets to offer the highest-risk and most popular form of sports betting: parlays. 

This would be a professionalized version of the “combos” that Kalshi already offers, so it seems likely to happen.

Prediction markets will be even more like gambling (and more negative-sum, probably) if they do.

Back to Satoshi

A New York Times journalist spent “more than a year digging into Satoshi’s identity” — and reported it as a personal memoir.  

This turns a pretty tired subject — who is Satoshi? — into an engaging read. But the reporter-as-protagonist framing also lowers the bar for what news is fit for The New York Times to print.

Here, the evidence the author collected is circumstantial and probabilistic: writing patterns, timing coincidences, ideological alignment. 

“He sometimes confused ‘it’s’ and 'its’,” the author says of both Adam Back and Satoshi.  They also both put two spaces after periods.  They both had an interest in Japan.

(Wait…am I Satoshi???)

Turning the reporting into a narrative also removes the need to consider disconfirming evidence. The author adds that both Back and Satoshi were versed in C++, but ignores how different their coding styles are. 

(I don’t write C++ at all, so I guess I’m not Satoshi. Damn.)

The author does report a moment of pause when he meets Mr. Back at a conference, “flanked by two executives from a new Bitcoin treasury company he had co-founded.”

For the CEO of a Bitcoin treasury company aiming to list on a US stock exchange, failing to disclose a holding of 1.1 million bitcoin would surely break some securities laws.

It’s also hard to imagine the real Satoshi partnering with Cantor Fitzgerald to put his coins on a stock exchange.

At any rate, as entertaining as the article is — “his shifty eyes, his awkward chuckle, the jerky movement of his left hand…struck me as fishy” — there’s no smoking gun here.

But there probably will be soon. 

Mythos will figure this out for us. 

Mimicry finance

An AI-driven study finds that asset managers are predictable: “71% of portfolio managers’ trades on average can be predicted in a given quarter given their past trading history.”

It also finds that predictable is bad: “Managers whose trading is the least predictable significantly outperform their peers, while those whose trading is most predictable significantly underperform their peers.” 

This runs counter to a lot of conventional wisdom in asset management. Typically, managers are encouraged to follow a consistent process, lean into styles like value or momentum, and repeatedly exploit whatever edge they think they have. But all that would presumably make a manager more predictable and, by extension, less profitable.

The same pattern holds among individual trades. “Even within each manager's portfolio,” the authors explain, “those stock positions that are more difficult to predict strongly outperform those that are easier to predict.” 

So you want to make your trades unpredictable.

Even better, you want to make unpredictable trades in unpredictable stocks: “When aggregating across managers, stocks in which the behavior of fund managers are least predictable strongly outperform stocks in which the behavior of fund managers are most predictable.”

I don’t know exactly how you’d do that. Maybe you run your typical investing process and then do the opposite of whatever conclusion you reach? Seems like that would work! 

Unless everyone starts doing it — in which case you’d have to do the opposite of the opposite. 

Widely adopted, that could reduce markets to a Sicilian battle of wits: I know that you know that I know…

The authors believe their study has implications beyond markets. “In theory, compensation to agents [such as fund managers] should not be tied to dimensions of behavior or human capital that can be replicated at low cost.” 

In other words, humans shouldn’t be paid much for behavior that an AI can mimic.

In both markets and life, it pays to be unpredictable. 

The coming shortage of humans?

In its paper refuting the doomer Citrini piece that recently went viral, Citadel included a reassuring note on human employment:  “If the marginal cost of compute rises above the marginal cost of human labor for certain tasks,” they wrote, “substitution will not occur.”

That point didn’t get much attention, likely because we assume AIs are cheap labor. But it’s already looking prescient because signs abound that we’re nearing a computing crunch that will only be solvable with higher prices. 

Anthropic is reportedly nerfing its Claude chatbot, presumably to ration scarce capacity. It also added usage limits for weekday afternoons, and forced users to pay-as-they-go for OpenClaw. OpenAI probably killed Sora for the same reason.

Importantly, all of this is happening before the release of Mythos, which is said to be significantly more compute-intensive. 

Stratechery’s Ben Thompson suspects Claude is delaying the public release of Mythos in part because it’s so compute-intensive they don’t have enough capacity to allow people to use it.

Similarly, engineer/writer Martin Alderson says “the next 18-24 months are going to be defined by compute shortages.”

If so, there might suddenly be human shortages, too.

— Byron Gilliam

Introducing Blockworks Investor Relations, an IR platform built for onchain businesses.

The latest Blockworks offering brings together analytics, a branded investor relations site, and integrated advisory support into a single platform. The result is a more efficient way to share your story, build trust with investors, and engage a global audience from day one.

Check out our cofounder Michael Ippolito's keynote at DAS NYC launching the new IR platform.