- The Skinny on Wall Street
- Posts
- The Skinny On...
The Skinny On...
Stable Savior | Stupid Bots
The Wall Street Skinny
August 01, 2025
This Week on The Floor
We are launching our first ever LinkedIn Live series! The subject?
Join us for a FREE 30-minute fireside chat where you’ll get insight from industry veterans and thought leaders, actionable steps, and strategic skills that will help you future-proof yourself professionally.
Registrants also get a 10% discount on our flagship Investment Banking & Private Equity Fundamentals self-paced course!
Make sure to follow us on LinkedIn for more exclusive and live content you won’t get anywhere else!
The GENIUS Act: save the stablecoins, save the bond market?!?!
Meme stock markets: what AI trading bots colluding means for investors, regulators, and the future of free markets
Markets Recap / Deal News
Interviewing this week? Here’s some content for your conversation.
Can Stablecoins Do What the Fed Won’t?
President Trump recently signed into law the GENIUS Act (Guiding and Establishing National Innovation for U.S. Stablecoins), marking the first major federal legislation regulating stablecoins in the U.S. markets. The bill passed with bipartisan support and has the potential to reshape not only the crypto space, but also the world of traditional finance.
The GENIUS Act creates a national framework for what it defines as “payment stablecoins” cryptocurrencies like USDC, Tether (USDT), and DAI. Stablecoins are exchange mechanisms whose value is not intended to fluctuate like that of a Bitcoin or Ethereum. Instead, they are typically pegged to a fiat currency (in this case, the U.S. dollar) and designed for transactional use in the crypto world. The law restricts stablecoin issuance to bank-affiliated entities or nonbanks approved by state or federal regulators. Issuers above a certain market cap must submit to federal oversight, while smaller players may remain under state supervision if they meet equivalent standards.
The Act also imposes new consumer protections. Stablecoins must be fully backed by low-risk, liquid assets like U.S. Treasury bills, and issuers are required to provide monthly reserve certifications and annual audits. Rehypothecation (the practice of reusing reserve assets for investment) is banned, and stablecoin issuers must comply with anti-money laundering and sanctions rules.
Additionally, the law establishes that stablecoin reserves are customer property in bankruptcy proceedings, ensuring that holders are repaid ahead of other creditors. Think of it as akin to the FDIC insurance policy on bank deposits. That was something that stablecoins lacked during previous collapses, with disastrous consequences for investors.
Beyond crypto policy, broader adoption of stablecoins could have real macroeconomic implications. As Treasury Secretary Scott Bessent recently noted, stablecoins may reach a $2 trillion market cap in the coming years. Because stablecoins are typically backed by short-term Treasuries, rising demand could create a persistent bid for t-bills. And when there’s more demand for t-bills, prices rise…and yields (interest rates) FALL.
Ironically, in a week where the Fed has kept rates on old yet again, despite pressure from pundits and presidents alike, we now have a new potential source of downward pressure on front end yields.
In a nutshell? Stablecoin usage could end up pushing Treasury yields lower and making government borrowing cheaper. So in the wildest twist of all, what once seemed like a systemic threat to the traditional finance world might actually turn out to be a boon to things like front end U.S. Treasuries!
What Happens When AI Traders Collude?
If “dumb” bots are working together to rig markets, what does that mean for smart investors?
A recent Wharton study is getting a lot of press for its recent findings about AI.
The study found that AI-driven trading algorithms can spontaneously learn to collude. Essentially, relatively unsophisticated AI was found cooperating to keep prices — and their trading profits — higher than a truly competitive market would otherwise allow. And they were doing so without any explicit agreement or communication.
The study identified two distinct forms of AI collusion:
The first was a price-trigger strategy where algorithms use price changes as signals to implicitly coordinate and keep prices high. Price-trigger collusion mainly occurs under specific conditions, such as when markets are relatively calm and have very little unpredictable trading noise, or when there are fewer well-informed traders, reflecting an information asymmetry and lack of efficiency.
The second was driven by an “over-pruning” bias in their learning process that makes all the algorithms become overly cautious and avoid aggressive moves, ending up effectively cooperating by default. In unstable or more competitive situations where the price trigger strategy fails, the AI traders still end up colluding through the over-pruning mechanism, because their learning process pushes them all toward the same safe, non-competitive strategy.
Why do we care? Well, any collusion, whether driven by AI or humans, hurts the market by reducing liquidity (making it harder or more costly to trade) and by making prices less informative about assets’ true value. Mispricings beget mispricings. More importantly, our society is predicated on the principles of free markets, and we ban anti-competitive and price fixing practices in many cases because they are in direct conflict with the values we hold dear.
The findings of this Wharton study pose a serious challenge for regulators and investors alike.
GIven the algorithms collude with no overt agreement or intent, existing screening mechanisms in place may not be able to pick up on the activity until it’s too late. Furthermore, regulators’ attempts to limit AI trading collusion could unintentionally backfire, as they might encourage AI systems to operate more stealthily.
For investors and other market participants, the implication is that AI-powered trading might not only lead to massively distorted prices, but might ultimately undermine the kind of free market activity that rewards fundamental analysis, competition, and excellence.
As our friend John Normand pointed out in his analysis, this is yet another example of the “alignment” problem we grapple with as AI becomes increasingly interwoven with our day to day businesses. Put simply, he says:
“Every investor explicitly values efficiency, one expression of which is return maximisation. They also implicitly value fairness, assuming they believe in some degree of regulation. The AI models described in the paper evidence a narrow form of the first value (profit maximisation), but not the second value (fairness).”
The real problem, as Normand suggests, is that the field of AI ethics is seriously lagging that of AI development, possibly setting us all up for catastrophe.