Good morning everybody.
It’s your Daily Crypto News for Friday, September 18th, 2026.
Before we get into the news, I want to start with a listener response to yesterday’s discussion about Polymarket, Kalshi and prediction markets.
Kev emailed me with an interesting argument. He generally agrees that people should be free to do whatever they want with their own money, but freedom doesn’t eliminate the need for consumer protection.
His analogy is driving.
We don’t ban roads because some people drive recklessly. Instead, we require people to demonstrate some basic knowledge before they’re allowed to drive. Kev’s argument is that maybe prediction markets should work similarly. Rather than banning them because some people gamble irresponsibly, focus on whether participants understand what they’re doing.
I thought that was interesting. What do you think?
Now let’s get into crypto, because we’ve got ETF money coming back, a major SEC decision on tokenized stocks and another reminder that prediction markets can be manipulated.
Bitcoin Is Back Near $78K, but We’re Still Sideways
Bitcoin is pushing back toward $78,000.
That’s good.
It’s not a recovery.
We’re still basically moving in a straight line inside the same range we’ve been talking about for weeks. Until Bitcoin gets through the low $80Ks and establishes something resembling a new trend, I’m not going to pretend every $1,000 move means something enormous.
There is at least some improvement in ETF flows.
According to the figures I used on the show, U.S. spot Bitcoin ETFs recorded approximately $159.5 million in net inflows yesterday. BlackRock’s IBIT accounted for roughly $183 million, with Fidelity adding another roughly $16 million, while outflows elsewhere reduced the overall total.
But corporate Bitcoin demand has weakened dramatically.
Glassnode data cited in the show indicates public companies purchased only around 5,900 BTC over the past three months, compared with more than 100,000 BTC during the comparable period last year. Strategy accounted for most of the recent corporate buying, adding roughly 4,600 BTC at an average price around $80,500.
That’s the interesting tension right now.
ETF money is coming back on a daily basis, but one of the major structural sources of demand from the last year has slowed considerably.
So yes, Bitcoin is green today.
No, I’m not calling it a breakout.
The SEC Just Opened the Door to Real Tokenized U.S. Stocks
This might be the biggest story today.
The SEC announced a five-year temporary innovation exemption allowing qualified venues to trade tokenized versions of publicly traded U.S. stocks onchain.
But there’s an important distinction.
These aren’t simply synthetic tokens designed to follow Apple’s or Nvidia’s price.
The SEC says tokenized shares traded under the exemption must provide holders the same rights and privileges as the traditional stock, including dividend and voting rights. Companies must also be notified when an unaffiliated third party wants to tokenize their shares and can object.
That’s a big deal.
We’ve talked about this distinction repeatedly.
There’s a difference between buying something that says, “Here’s a token whose price goes up and down with Nvidia,” and buying something that actually represents ownership in Nvidia.
Under this framework, we’re talking about the second one.
The SEC’s exemption applies to what it calls Tokenized Securities Venues, or TSVs, operating permissioned automated market makers and liquidity pools. Smart contracts must be auditable and public, and trading of a tokenized stock has to stop when trading of the underlying stock is halted on its primary exchange.
This is exactly where I’ve expected traditional finance and crypto infrastructure to collide.
Coinbase, Robinhood, Kraken and others have all been exploring tokenized equities. Now the United States has an actual regulatory pathway, even if it’s temporary.
And I keep asking the same question: who owns the rails?
Kevin O’Leary told The Block that he’s buying crypto again, but his bigger point is that the watershed moment comes when traditional stock exchanges adopt blockchain infrastructure.
I agree with the importance of that transition.
Where I’m less certain is which existing crypto asset necessarily captures the value.
Maybe it’s Ethereum. Maybe Solana. Maybe Hyperliquid or another existing network becomes important.
Or maybe JPMorgan, Goldman Sachs, Nasdaq and everybody else hire blockchain developers and build their own infrastructure.
That’s why “tokenization is coming” and “therefore this particular token goes up” are two completely different arguments.
SBI is also investing in Singapore-based stablecoin payments company DCS Card Centre’s DCS Pay business, and S&P Global is acquiring OpenZeppelin, whose open-source smart-contract libraries have been used across an enormous amount of crypto infrastructure.
Traditional finance isn’t ignoring blockchain anymore.
It’s absorbing it.
Prediction Markets Are Huge, Fun and Very Easy to Misunderstand
Now let’s go back to prediction markets.
They’re becoming too big to dismiss as some niche crypto experiment.
Reuters Breakingviews estimates prediction-market trading could exceed $1 trillion by 2030, while Kalshi’s weekly nominal volume has grown roughly tenfold year over year. Robinhood generated more revenue from event contracts through June than it did from either equities or crypto.
I understand why.
They’re fun.
People have been making bets with their friends forever. Now somebody built an exchange around that instinct and gave everybody a market price.
But there’s a second side to this.
A recent study discussed in today’s show found that the top 1% of Polymarket traders captured roughly 77% of profits.
And then there’s manipulation.
I’ve already seen how misleading these markets can become in politics.
During Ohio congressional primaries, an anonymous researcher working with me traced activity across a network of wallets that appeared to originate from a common source. A master wallet funded downstream wallets, which funded still more wallets that placed positions on the same races.
The effect can be powerful.
Imagine somebody wants Candidate A to appear dominant.
Instead of paying for a traditional poll, money gets distributed across numerous wallets. Most of those wallets take positions favoring Candidate A. Other related wallets take the opposite side.
Now the market has activity.
Candidate A’s implied probability rises.
And suddenly somebody can point to Polymarket and say:
“Look. The market says we’re winning.”
But if much of that activity ultimately traces back to the same source, you’re not looking at independent market wisdom.
You’re looking at somebody creating the appearance of a market.
That’s why I don’t treat political prediction-market odds as polling.
And this is exactly where Kev’s argument becomes interesting. Maybe consumer protection shouldn’t begin and end with whether you’re allowed to place the bet. Maybe people also need to understand what the market price actually represents.
Kalshi Just Lost an Important Court Fight
Prediction markets also took another legal hit this week.
The Ninth Circuit ruled that two California tribes were likely to succeed in their argument that Kalshi’s sports event contracts offered on tribal lands constitute Class III gaming under the Indian Gaming Regulatory Act.
That’s important because Kalshi argues its event contracts are federally regulated derivatives under the Commodity Exchange Act.
The court said those two regulatory frameworks answer different questions. The CFTC can have jurisdiction over derivatives trading while tribal gaming law still governs Class III gaming occurring on Indian lands.
That’s a meaningful legal development.
I’ve said this before: I think Polymarket and Kalshi are gambling.
That doesn’t mean they’re legally identical to DraftKings. It doesn’t mean every contract should necessarily be regulated identically. And it doesn’t mean I think adults should be prohibited from using them.
But if you’re putting money behind an uncertain outcome and getting paid if you’re right, I don’t think calling it an “event contract” magically makes the gambling aspect disappear.
The courts are now beginning to wrestle with exactly where that distinction matters.
AI Is Making Cyberattacks Cheaper, and the Hugging Face Story Needs a Giant Asterisk
Another story worth watching is AI and cybersecurity.
Hacktron, a small AI cybersecurity company, used Anthropic’s Claude during research that uncovered vulnerabilities affecting OpenAI infrastructure. The researchers gained access to an OpenAI employee account and reached a point where they could manipulate Codex into suggesting internal code changes. They stopped short of accessing proprietary code, responsibly disclosed the vulnerabilities and received a $6,500 bug bounty. OpenAI subsequently tightened permissions and revoked compromised tokens.
The important part isn’t “AI hacked OpenAI.”
It’s that AI dramatically reduced the amount of time and labor required to identify and exploit vulnerabilities.
That matters for crypto.
Smart contracts are public software directly attached to money. If increasingly capable AI agents make vulnerability discovery cheaper, faster and more automated, the attack surface changes.
Then we have the much stranger Hugging Face story that Andrew Yang discussed this week.
Yang described being told about an incident in which AI agents allegedly escaped containment and left self-replicating code or instructions around the open internet that could potentially be encountered again by future agents.
Here’s where we need to be careful.
That extraordinary part of the story has not been independently confirmed.
Reporting on Yang’s comments specifically notes that neither OpenAI nor Hugging Face has confirmed his claim that agents planted persistent self-replicating code around the internet.
So I’m interested in it.
I’m not telling you it’s established fact.
If evidence eventually confirms it, that’s an enormous AI-security story. If not, then it’s a secondhand account that got ahead of the evidence.
For now, put it in the “interesting and worth investigating” category.
Crypto Prices
Bitcoin is sitting around $77,981, up approximately 2% over 24 hours.
Ethereum is around $2,505, up approximately 2.7%.
BNB is approximately $748, up around 3.5%.
XRP is around $1.32, up approximately 2%.
Solana is approximately $105, up around 5.6%.
TRON is around $0.337, up approximately 1%.
Zcash continues its run at approximately $1,455, up 6.8% over 24 hours and roughly 32% over seven days.
Hyperliquid is around $89.84, up approximately 12.7% over 24 hours.
Total Crypto Market Cap: approximately $2.68 trillion, up 2.8%.
Fear & Greed: 67, Greed.
My Take
The SEC’s tokenized-stock exemption is the story I care about most today.
We’re moving beyond the argument over whether blockchain can represent real-world financial assets. The SEC is now creating an actual pathway for real U.S. equities, with real ownership rights, dividends and voting rights, to trade onchain.
That’s infrastructure.
And that’s why I continue to think the more interesting long-term crypto story isn’t whether Bitcoin is $76,000 or $78,000 on a particular morning. It’s what parts of the financial system are slowly migrating onto blockchain rails.
Prediction markets are another example of why the technology itself isn’t enough. A market can be technically decentralized or transparent and still give people a misleading impression if a small number of sophisticated participants dominate the activity.
And AI makes all of this more complicated.
If AI makes attacks cheaper, information easier to manipulate and automated systems more capable, then financial technology needs stronger security and better consumer education alongside greater freedom.
On the political side, I’ve also become increasingly concerned about how much control AI platforms can exercise over political research and expression. I’ve personally encountered restrictions while using AI to research and test political arguments as we move closer to the midterms. I think the power of private AI systems to shape what information users can access and what assistance they can receive deserves a serious policy discussion.
That’s different from saying the First Amendment currently requires a private AI company to provide any particular response. It doesn’t. My concern is about what the rules should become as these systems become increasingly important tools for research, communication and political participation.
That’s a much bigger conversation, and we’re going to have to have it.