🧠 Binance hacked for $6.36B. Then came the freeze.

My guys at Modern Tresury dropping knowledge 👇️

In lending, real estate, and personal finance, payment mistakes don’t stay on a dashboard. They show up in someone’s home, savings, or paycheck.

So how do the teams who get it right actually do it?

Modern Treasury is bringing leaders from three of them together for a live session, How They Built It:

• NFTYDoor funds home equity loans 24/7.

• GrowIt grew real estate investing from zero to $175M in equity under management.

• Albert runs personal finance for 10M+ people.

Hear which rails they chose, how they track every dollar, what changed as they scaled, and what they’d do differently next time.

Don’t miss their takeaways: RSVP here

Things to know 👀

BONUS: Nerdcon announces MoneyCon. For everyone who loved coming to Vegas to network and talking payments and finance, meeting with entrepreneurs, founders and getting business done. We’re bringing back restaurant row. Get in touch to learn more. Discounts available for any Nerdcon sponsors or ticket holders.

1. Binance lost $6.36B. Issuers froze about 78% of the stablecoins.

It's the biggest crypto theft on record, about four times the $1.5B Bybit hack in February 2025. TRM Labs attributes it with high confidence to TraderTraitor, the North Korean unit the FBI blamed for Bybit. Most of the money left Binance wallets on ten blockchains in about 20 minutes, and Binance says the attacker got in through a compromised engineer's laptop and used that access to submit withdrawals.

Sometimes a tweet says everything you want to say:

🧠 Stablecoins can be frozen quickly and that’s really something. About $3.4B of the haul was in stablecoins such as USDT, USDC and RLUSD. Tether, Circle, Ripple and World Liberty Financial froze about $2.67B of it, and Tether's first freeze came 37 minutes after the biggest transfer. No issuer can freeze ETH or BTC, and the attacker still holds about $2.4B of them.

🧠 The attacker raced the freezes. It sold 325M USDT to market makers for ETH and paid more than 10% over the market price to do it. Another wallet deposited 60M USDT on the lending app Aave and borrowed $44.6M of ETH about 40 minutes before the first freeze.

🧠 The asset you hold in a hot wallet decides how much you can claw back after a breach. Most of the stolen stablecoins got frozen, and about 1% of everything else did.

🧠 For everyone in crypto, we need to be counterparty screening in real-time. TRM says screening would have flagged a wallet selling USDT it had received from the hack minutes earlier. Yes, they would say that since that’s what they do, but it works.

🧠 A freeze stops the money moving, and getting it back usually needs a court order. Tether has burned frozen USDT and reissued it to government wallets before, and recovery can take months or years.

🧠 Users don’t have to wait for the court order. Binance says it will cover all losses, and Fundstrat doesn't see a solvency risk. Binance has been for the past few years one of the most ludicrously profitable businesses on earth so no doubt they can afford this.

🧠 Enterprises who are adopting stablecoins woke up today asking very fair questions of their entire supplier ecosystem. How do you know this won’t happen to me? Who are your third parties and is their code secure? Are your companies laptops secure. Finance is hard people.

🧠 North Korea has been a consistently capable hacker, and you have to wonder if AI has accelerated their ability to find vulnerabilities. As crypto and digital assets go mainstream the need for rock solid cybersecurity has never been higher. It’s time to get deadly serious about getting this right.

2. Kalshi raises $1B at a $40B valuation

Kalshi, the US prediction market, raised $1B at a $40B valuation in a round co-led by Sequoia and Wellington. It was worth $5B a year ago and $22B in May. A day before the announcement, Utah filed 14 criminal charges against Kalshi and co-founders Tarek Mansour and Luana Lopes Lara, accusing them of promoting illegal gambling.

🧠 Sports brought in most of an estimated $4B revenue run rate in July, but Kalshi says prediction markets are only its first step toward a full exchange. The new money goes into leveraged event contracts, perps (futures that never expire) on stocks like Tesla and Nvidia, and markets that trade the cost of AI computing power. That puts Kalshi up against CME and ICE, which owns the New York Stock Exchange.

🧠 All that sports volume carries massive consumer harm risks. Mansour told Bloomberg this week that irresponsible users are "a real worry," and Kalshi set a default $2,500 monthly deposit limit for under-21s. Its gold perp already offers up to 15x leverage.

🧠 Mansour says Kalshi takes a 1% fee whether you win or lose, so it can't be the house. Institutions make up a quarter of Kalshi's volume before you count market makers like Susquehanna, and last week Utah sued two Susquehanna firms for allegedly taking bets against users. The house is now Wall St.

🧠 Enforcing these protections in a big way now would be bad for business. A 1% fee is something a consumer will pay, but at scale in financial markets that's a ludicrous price.

🧠 Few things can be true at once (per Pet Berisha of Predicted):

1. “Kalshi & PM want to do more Tradfi things and operate in traditional/frontier capital markets.

2. Every single quarter since we've tracked data, Kalshi has become MORE sports-centric, not less

3. In Q3, 44% of Kalshi's card deposits came from the 11 states that do not allow sports gambling…”

[Data sourced from Americanspend]

🧠 I'm glad we're seeing these steps towards consumer protection, and it's still progress. But I feel like these companies really need someone to tell them to protect consumers, because it's too profitable to have them as the marks.

🧠 The CFTC is actively looking at what applies here. It proposed rules for prediction markets in June, and on Tuesday it sued Utah to stop the state applying gambling law to federal exchanges.

🧠 Two appeals courts say states can regulate sports contracts and a third disagrees, so I think Kalshi's strategy is to go all the way to the Supreme Court. If it wins, CME and ICE get a 24/7 competitor with millions of retail customers. If it loses, states get to regulate the sports contracts that bring in most of its money.

3. Instinct launched Balance, and gave AI agents a stablecoin wallet

Instinct users can now load a balance, set a limit per purchase, and let their agent buy things without approving each one. The demo shows someone adding $700, asking for their favourite lunch every day for a week, then concert tickets. Each agent holds OUSD (a stablecoin) in a wallet from Privy on Tempo*, and the purchase still runs on Visa.

🧠 The card networks are not cooked. For a $50 purchase, Stripe sells $50 of stablecoin, the dollars fund a prepaid card account at Lead Bank, and the purchase runs through Visa. The merchant gets paid the way it always does and needs no wallet.

🧠 Link balance is the consumer’s control panel for the agent. Tom Noyes points out that prefunding it from a linked debit card does two jobs at once: the money is already there, so the authorization problem goes away, and the agent’s card runs on cheaper funds than a commercial card at ~300bps. Stripe and the agent platform both make money on that structure.

🧠 The balance also earns yield for Stripe or its partners while it waits to be spent, instead of sitting at a bank earning nothing.

🧠 A wallet funds more than one card. One agent balance can fund many cards, plus wire, ACH and stablecoin payments, and once enough users hold one, agents can pay other agents or any stablecoin endpoint directly.

🧠 The bigger story is Stripe building a consumer business. Tom’s estimate is that Stripe sits under roughly 90% of US and EU agentic transactions today (ACP with Shopify, and Link behind Muse and Instinct), because right now it’s the only company that can make the whole flow work.

🧠 The card networks and merchants are building their own alternatives to this balance approach to not having to auth every transaction. Walmart’s Agent Gateway (21 Oct) keeps fraud and risk with the merchant, who decides which agents get in and lets the customer set a spending limit. Amex’s ACE has the consumer register intent and permissions with the card itself, and Visa and Mastercard have similar standards, though issuers won’t let the networks move into authentication and blocking.

🧠 Everyday spending is the consumer wedge for AI. Cash App, Chime and Revolut peeled off everyday spend first and won the paycheck later, and consumer agents are starting in the same place.

🧠 Zuckerberg said shopping is Muse’s most common use case, with an average purchase over $200. Giving an agent a balance is the first step toward it managing a budget, and then money more broadly.

4. Ramp opens Inspect, the coding agent behind 74% of its merged PRs

Ramp opened early access to Inspect, its internal coding agent, on Tuesday. You tag @inspect in Slack with a screenshot of a bug, and it gathers context, fixes it, attaches visual proof and prepares a pull request for review. In the past month, 99% of Ramp's engineering, design, and product teams used it; it now accounts for 74% of merged PRs, and Ramp says its builders ship 6x as much code per person as a year ago.

🧠 Having an audience for AI being used is underrated. Nobody mandated it. Every session is multiplayer by default, so one person fixing a bug in Slack is a live demo for everyone else in the channel.

🧠 74% is class-leading. I hear 40% and 50% from the companies that talk about this at all. And while code volume isn’t a proxy for quality, 6x more of anything if the quality bar is even half as good is still a net productivity gain.

🧠 Nugget from the Eric interview: Ramp was one of the first customers of Claude Code. By constantly pushing the boundaries of what the tool could do, they found the workflows that became Inspect.

🧠 Ramp's internal tooling is becoming productivity software. Router.com shipped in August, Glass (an AI workspace) on 13 October, and Inspect (also with a shiny .com domain) today. Rather than having employees spin out and build new companies, Ramp builds products that seem miles away from fintech.

🧠 It's weirdly mission-aligned. Eric Glyman told me in New York last week: "We're more of a productivity company than a finance company, save time + money." It's easier to save money if you can move it, and better workflows are another way to save it.

🧠 Two weeks ago Ramp raised $1.85B at a $60B valuation. I think investors are pricing the productivity company, and the corporate card comes with it.

5. DTCC's tokenization service goes live as 33 companies block OKXICE

DTCC, which holds $114 trillion of securities for US banks and brokers, launched a service yesterday that turns those holdings into tokens on a blockchain. The same week, 33 companies including JPMorgan, Microsoft and Goldman Sachs objected to their shares trading on OKXICE, the 24/7 tokenized stock venue from OKX and NYSE owner ICE.

🧠 DTCC's tokenization service is the biggest thing to happen in tokenization this year. Russell 1000 companies can now have a digital twin of their stock created, starting with a first batch of 34 securities, if the banks and brokers opt in to using it.

🧠 DTCC's tokens move between registered wallets 24/7 from day one, with 20 firms signed up including JPMorgan, Citi and Fidelity. Next comes tokenized collateral, so banks can move it on a weekend.

🧠 Why wouldn't every neobank and brokerage want this, especially if it reduces their collateral requirements? Companies like Alpaca are doing their best to become the "Plaid of tokenized stocks" and deliver this capability everywhere.

🧠 Remember what happened to Robinhood in 2021 during memestock mania? Extreme retail trading in stocks like GameStop and AMC led DTCC's subsidiary, the National Securities Clearing Corporation (NSCC), to ask Robinhood for $3bn at 3:30am to cover trades still waiting to settle.

🧠 Tokenized stocks can settle the moment they trade, with the stock and the cash swapping hands at once, so no trades sit waiting and NSCC has nothing to collect cash against. Vlad Tenev made this argument in January. For now, exchange trades still settle in one day even in tokenized form, and settling against digital cash is on DTCC’s roadmap.

🧠 However, the alternative models are struggling. OKXICE uses the SEC's Innovation Exemption to trade tokenized stocks, but that exemption gives each company a veto, and about half the companies on its list have used it.

🧠 Some of the banks that designed the DTCC service are the same ones objecting to OKXICE. This is Wall Street choosing its own rails, but with some good reasons. The OKXICE venue sits outside Reg NMS, the rules that make sure investors get the best displayed price, its liquidity providers can leave at any time, and it's unclear whether self-custodied tokens are protected if the tokenizer's broker fails.

🧠 The US now has three kinds of tokenized stock.

• DTCC's tokens need banks and brokers to opt in.

• OKXICE's tokens need each company to not object for 30 days.

• Offshore tokens like Kraken's xStocks give non-US users price exposure only.

🧠 Companies only get a veto over the middle one, and they're using it. Objecting costs nothing and silence can't be undone, so legal teams say no first and study it later. AMC's CEO famously took to X to complain about the offshore model.

🧠 You have to imagine the DTCC model is the one with legs, at least domestically, and someone like OKXICE could adopt it quite quickly. We will get 24/7 trading, but some of the experimental stuff might fall away before then.

🧠 There’s also secret option 4, listing your stock directly onchain with the secretary of state. So for example, a Delaware company could tell their state secretary to list, and directly issue, and create its own functionality around that. Time will tell if that catches on.

🧠 I'm excited about the idea of tokenized stocks expanding the TAM and functionality of stocks. Tokens can move 24/7 and globally much more easily, meaning the global south could become holders. You can bake in things like voting as a simple bit of functionality. And we're already seeing folks use stocks as collateral for lending. Imagine if a stock portfolio could help finance someone's first home?

4 Companies 💸

1. Beltic - Know Your Agent checks for merchants and payment companies

Beltic lets merchants, payment companies, and banks see which AI agents are hitting their site or API, who operates each one, and what it's trying to do. Teams write rules for what agents are allowed to do, like confirming multi-bookings from new accounts or blocking high-risk actions automatically. Agents it can't verify get flagged.

🧠 Right now, an AI agent logs in with your password and looks exactly like you. The merchant can't tell whether it's you, your assistant, or a fraudster's bot, so the safe move is to block all of it. Beltic lets them say yes to the good ones, and starting with the fraud team feels right because they have the problem today. But how do you win distribution here? Visa, Mastercard, and Cloudflare are all shipping their own agent verification, so does a merchant buy this from a startup?

2. Catalyst - The prosumer investor's AI

Catalyst turns a trading idea typed in plain text that watches the markets and places the trades. It covers crypto tokens, perps, tokenized stocks, and prediction markets, with signals from news, social activity, and sentiment. Users keep custody se or change a strategy at any time.

🧠This struck me as for that guy with 5 screens in his mom's basement stitching together alerts, bots, and a tab for each exchange. Typing the trade in a sentence and letting software watch for it is an easy sell. The company says a pilot with power traders did hundreds of millions of volume in about two weeks. The plan is to take it to everyday investors next. But does the everyday investor know what to do with a plain text box in front of them? Or do they have the drop-off problem a lot of AI products have in mainstream?

3. Peppermint - Billing and collections for clinical trial sites

Peppermint helps clinical research sites get paid for the trials they run. It pulls visits and procedures from the site's trial software, checks each chart budget, then sends the invoice, chases the payment, and matches the cash. Sites can hand over billing entirely or use it alongside their own team.

🧠 Trial sites get paid per patient visit and per procedure, under a contract negotiated study by study. Working out what they're owed means stitching it to mint claims it finds 15% in missed revenue. It's revenue cycle management for trials, and selling it as a done-for-you service is smart, because a small research site wants the cash collected. I d is before they have to sell to hospitals and sponsors.

4. IMMIX - TP-ICAP for Stablecoins.

IMMIX gives payment companies, stablecoin issuers and exchanges a price to swap between non-dollar stablecoins and currencies. It quotes one firm price and then trades itself, so a payment settles as a single FX transaction. The company says it has handled more than $25 billion in customer volume since 2024.

🧠 TP-ICAP is the world's largest FX brokerage. There isn’t yet an equivalent for stablecoins (although OpenFX and XFX are gunning for that spot). FX in stablecoins is often poor compared to wholesale. Outside the dollar, liquidity is so thin that, per IMMIX, one conversion can break into three or four separate trades, and you pay a spread on each. In wholesale FX, a bank quotes you one price, and IMMIX wants to be that dealer for stablecoins. All of these companies are as good as the liquidity they can build. But this is a London-based team with FX heritage. So let's see.

Good Reads 📚

1. The commanding heights

This incredibly long piece took me about an hour to read, is the talk of the AI and Wall St crowd and essentially argues.

• The US and China will be the two countries with AI, which creates a cold war condition.

• The US's durable advantage is compute and capital, with 70% of compute compared to 5% in China, meaning they can stay at the frontier.

• The US state will see this as a coming cold war, and look to own more of these companies, just as threatened states have historically in war conditions

• China will not monetize its models because it won’t let its currency float, meaning even Chinese models will be better monetized by the dollar.

• The most neutral dollar is a stablecoin, which, while not always controlled by the US as tightly as the fiat dollar, will enable global debasement

• AI labs will make such a massive portion of the K-shaped economy; the government needs them to continue to grow

• Therefore, the capex must continue to go up, and the US Government will backstop that and, potentially, put a limit on the yield from US Treasuries to ensure funding can continue into capex.

• Therefore, the US government looks to own these companies either via a golden share or gifted equity.

• AI will inextricably replace labor, causing developing economies to enter economic hardship, but the US will be partly insulated by government ownership enabling some level of redistribution

• The value of AI will not accrue to labs, which will become government-subsidized utilities, or the existing AI supply chain bottlenecks, but to the other industries that emerge because AI exists.

🧠 This is dystopian as hell. And while the chain of logic is rational, it's also not confidence-weighted. There are several thousand words devoted to Ravi, who works in India, who’s basically seeing all of his friends out of work and is ultimately laid off via WhatsApp.

🧠 They call the Wall St leveraged buyout firms of the 1980s “buyout artists.” This is clearly a play to attract capital to raise a sort of permanent capita / hedge fund vehicle for buyouts.

🧠 The link for stablecoins is one of the least worst I’ve seen to AI. Person in Argentina wants to get paid and pay for AI, stablecoin linked cards are a great way to do that.

🧠 People are saying this is bullish for certain stocks, while hyperscaler share prices are cratering. As I write this, Amazon stock is down 15% over the week, and Google is down 16.7%. That could be the essay’s thesis, that value is moving away from those who own the rails. But its more about these companies entering negative free cash flow and AI impacting their ads business.

🧠 I don’t buy that existing incumbents are sitting ducks here. Regulation exists, network effects exist. Someone like Amazon or JP Morgan can block AI agents, and will do so, and will bring people to the negotiating table.

2. IMF study of stablecoins in 188 markets.

The IMF constructed a globally representative panel of money flows in 188 countries representing more than $21.4 trillion in gross flows from 2018 to 2025. What they found is fascinating

Activity is disproportionately concentrated in emerging markets and developing economies, which hold three quarters of stablecoins. When countries restrict capital outflows, stablecoin activity rises by roughly 30 percent over the following six months, then plateaus, and does not fall back when restrictions are removed. When currency crises hit, stablecoin holdings increase by roughly 75 percent within a year, and the new holdings persist.

🧠 Stablecoins are the dollarization of choice

Tweets of the week 🕊

That's all, folks. 👋

Remember, if you're enjoying this content, please do tell all your fintech friends to check it out and hit the subscribe button :)

Want more? I also run the Tokenized podcast and newsletter.

(1) All content and views expressed here are the authors' personal opinions and do not reflect the views of any of their employers or employees.

(2) All companies or assets mentioned by the author in which the author has a personal and/or financial interest are denoted with a *. None of the above constitutes investment advice, and you should seek independent advice before making any investment decisions.

(3) Any companies mentioned are top of mind and used for illustrative purposes only.

(4) A team of researchers has not rigorously fact-checked this. Please don't take it as gospel—strong opinions weakly held

(5) Citations may be missing, and I’ve done my best to cite, but I will always aim to update and correct the live version where possible. If I cited you and got the referencing wrong, please reach out