The Chopping Block: The Binance Hack, Circle's Fast Freeze, YC's Crypto Wishlist, and ICE Bringing Hyperliquid Perps Onshore

The hosts had planned to avoid another discussion of crypto hacks, but roughly three hours before recording, Binance appeared to suffer what Tom described as the largest hack in crypto history by a wide margin. Early estimates put the loss at about $5 billion—roughly three times the previous record—with slightly more than half in stablecoins, primarily Tether and some USDC. Circle and Tether appeared to freeze the identifiable stablecoin funds quickly, while other assets were being swapped, moved into Tornado Cash, or otherwise exfiltrated. Tom stressed that the situation was still developing and that everyone was operating in the “fog of war.” Tom’s preliminary technical explanation focused on Binance’s handling of EIP-7702 delegation. The mechanism allows an address to delegate signature verification to another contract, so the delegated signer can move assets or call code on the user’s behalf. His theory was that an attacker exploited the way Binance managed that delegation, generated signatures, and withdrew funds by proxy. Although the loss was enormous, Tom said Binance had enough balance-sheet strength to make customers whole; CZ publicly indicated that funds were safe. The group nevertheless viewed the event as another escalation in the security stakes as more activity moves on-chain.

The discussion turned to Circle and Tether’s response. Circle is often criticized for moving slowly when asked to freeze funds, because it generally wants a court order or law-enforcement involvement, while Tether has historically been seen as more willing to act on its own judgment. This time, however, both reacted quickly, with Circle freezing USDC associated with the incident almost immediately. The hosts joked that Circle was finally moving fast—“Not today, though. Not today”—and noted the apparent contradiction between criticizing a freeze when it is slow and criticizing it when it happens quickly. Nemil, who previously worked at Coinbase on USDC, said the Coinbase and Circle teams had literally practiced scenarios like this. Their playbook could include pausing the stablecoin’s smart-contract functionality, blacklisting particular funds, or temporarily preventing transfers while the facts were investigated. He compared the problem to social-media takedown requests: Circle receives huge numbers of claims from people saying a wallet was drained, and it cannot reliably adjudicate every one. A direct, verifiable report from Binance is much higher signal. The group also emphasized that freezing fungible dollars can disrupt DeFi protocols, since stolen stablecoins may be deposited, leveraged, or used as collateral before anyone can determine whether the freeze request is legitimate.

Robert argued that DeFi still has not fully solved the problem of assets being frozen by an outside issuer. Four or four and a half years after the failures of 2022, entrepreneurs have had plenty of time to propose new architectures involving custodians, triparty arrangements, or different forms of segregation, but he could not name a mainstream system that had genuinely shipped and gained adoption. DeFi protocols increasingly have protocol-level emergency switches, and several reportedly paused new activity soon after the Binance incident, but individual-asset freezes remain difficult for protocols to handle. The participants considered whether the industry is moving toward a TradFi-like model in which most exchange assets are held through triparty custody rather than directly in smart contracts. Such arrangements can reduce counterparty and custody risk, but they also impose a cost of capital and make assets less immediately productive in DeFi. Nemil said centralization increasingly functions as a security or “break glass” measure for stablecoins, tokenized stocks, and other regulated assets, especially as AI enables more sophisticated attacks. Rob noted that many of his stablecoin and institutional-infrastructure portfolio companies were not directly affected because their businesses already incorporated centralized controls. Nemil agreed that crypto is more mature, but warned that financial crises are driven by interconnection: if Binance could not cover the loss, the result could have been a bank-run-like cascade similar to FTX.

Nemil explained why he remains optimistic despite the dejection he sees among people who entered crypto for price appreciation. In his view, the last six months have delivered the institutional adoption he hoped for a decade ago: major banks, payments companies, and other incumbents are seriously considering crypto integration. He contrasted that with 2021’s speculative activity, including Olympus DAO forks, which created what he called false product-market fit. The meaningful question now is whether people want and use the products, not whether their tokens are rising. He outlined YC’s crypto requests for startups. The first is broader tokenization: stablecoins were the initial wave, but tokenized equities and other real-world assets are now developing. The second is creating assets and financial products that were not possible in the old system, such as micro-securities, micro-tokens, and smart contracts that let AI agents transact with one another. The third is building financial protocols, including regulated or centralized finance protocols with privileged operators. Once those systems exist as APIs, agents can use them for payments, trading, and other transactions. Nemil’s final major theme was on-chain capital formation for AI. With hundreds of billions of dollars being spent annually on AI infrastructure, GPUs, and energy, he thinks blockchains could help raise and allocate capital for that buildout, much as crypto once enabled global fundraising during the ICO era.

Rob characterized the current period as “the bull run, but not the bull run the crypto people want”: institutional adoption, real businesses, and technical building are accelerating even while crypto social media feels disappointed because token prices are not providing the familiar feedback. Nemil said the earliest crypto vision was broader and more radical—potentially decoupling money from the state—so some early participants may see today’s institutional integration as less exciting even though it represents substantial progress. He tracks basic adoption metrics such as stablecoin market capitalization and transfer volume rather than treating market price as the health of the industry. The group also debated whether companies should be built before or after an idea becomes fashionable. Nemil said the best founders often work on non-consensus ideas when attention is elsewhere, citing Coinbase’s founding in 2012 and OpenAI’s early period as examples. Tom added that crypto repeatedly revisits ideas first attempted during the BitcoinTalk, BitShares, or Waves eras; later success often depends on a missing ingredient such as better technology, liquidity, regulation, or capital. Robert’s compressed verdict was that crypto is “post-cool, which also makes it pre-cool again.” The participants concluded that timing remains difficult, but founders should focus on a real unmet need rather than trying to predict which narrative will be hottest.

Asked whether crypto will mainly become infrastructure for AI, Nemil described a mixed future. YC had worked on X402, a system for agentic payments using stablecoins, but he has become less absolutist: credit cards are already easy to use and widely accepted, so every personal AI agent may simply have a credit card associated with it. Stablecoins become more compelling where they are materially better—for global payments, micropayments, or transactions that do not fit existing banking rails. He gave Cloudflare’s pay-to-crawl model as an example. A website could require an AI agent to pay before receiving content, and stablecoins may be particularly useful for that kind of global, small-value transaction. He also recalled that early stablecoin adoption was driven by access to crypto exchanges and DeFi, where a fiat dollar could not be used directly. The same test should apply to tokenized stocks and other assets: global accessibility, weekend trading, or another clear advantage must justify putting the asset on-chain. Nemil said the Iran-war-related demand to trade oil over a weekend illustrated how a specific event can suddenly make an on-chain market valuable. He was excited about regulated options and tokenized assets, but cautioned that institutions will require transparent custody, counterparty chains, and trustworthy structures rather than opaque vehicles.

The hosts then discussed what had been expected to be the day’s biggest announcement: ICE, the parent company of the New York Stock Exchange and a major derivatives exchange, signed a definitive agreement with trade.xyz to bring regulated perpetual futures to U.S. customers, subject to CFTC approval. Trade.xyz is associated with the HIP3 markets on Hyperliquid, including real-world-asset markets for oil and equities. The likely structure, though not explicitly confirmed, is that trade.xyz would pursue an FCM application, use Hyperliquid technology through a segregated “HIP3 Star” environment, apply KYC and reporting requirements, and potentially clear trades through ICE’s clearing infrastructure. The participants saw this as a major change from the era of private-chain pilots and proof-of-concepts. It could preserve the advantages of perpetuals—continuous trading without contract rolls—while adding the institutional protections of a regulated intermediary and clearinghouse. They also mentioned Stan Druckenmiller’s public interest in Hyperliquid and joked that he might not yet know what Hyperliquid’s auto-deleveraging, or ADL, mechanism is, though he would probably dislike it once explained. The group expected the regulated and offshore venues to remain separate rather than sharing one aggregated order book, with arbitrageurs and overlapping liquidity providers creating indirect connections.

Robert argued that perpetuals are a superior product format to dated futures because they do not expire or need to be rolled and can trade continuously, potentially including 24/7 markets. The obstacle is not the basic trading experience but institutional plumbing. Perpetuals require mechanisms such as funding payments every eight hours, liquidations, and sometimes ADL, while institutions want predictable credit exposure and do not want to be forcibly unwound because a counterparty or venue fails. Tom noted that the funding-rate machinery may be difficult for traditional exchanges and clearinghouses to reproduce, while institutions dislike the same features that retail traders accept. The group contrasted this with the crypto market’s earlier indifference to counterparty risk. Tom recalled that even a major prime broker such as FalconX had not previously completed a triparty transaction when he conducted what he believed was its first, whereas traditional hedge funds have long focused on custody, collateral, and counterparty exposure. FTX reinforced the lesson, and today more crypto businesses build in assumptions that venues can be hacked or liabilities can emerge. The conversation ended by framing crypto as too broad to treat as one industry: its subfields now range from Byzantine fault tolerance and blockchain performance to stablecoins, tokenized securities, credit, clearing, and AI agents. Nemil said people can still work on the underlying protocol problems, but the ecosystem no longer has a single center of gravity.