On February 28th around 6:14 AM UTC, the United States began bombing Iran jointly with Israel. Oil should have spiked. Gold should have ripped. Equity futures should have dumped. But the CME had been closed for eight hours. The NYSE wouldn't open for another 46. The most sophisticated capital markets in the world had nothing to say about the start of a war.
Meanwhile, on a decentralized exchange that most of Wall Street has never heard of, crude oil perps moved within seconds. Gold followed. A full grid of equity futures (Tesla, Nvidia, the Nasdaq 100) repriced within minutes. And by the time the CME reopened on Sunday evening, those prices turned out to be right. 100% directional accuracy across 35 instruments. Median error at the open: 14 basis points.
The Internet Financial System isn't a pitch deck anymore. It's a live market. And it's faster than yours.
@TheiaResearch wrote the most ambitious articulation of this idea last January: a vision for a global, permissionless, 24/7 financial infrastructure built on public blockchains. His argument: the existing financial system operates on 90,000+ siloed servers that can't talk to each other, charges enormous rents to gatekeep access, and arbitrarily shuts down for nights, weekends, and holidays. The Internet Financial System replaces all of that with a single open ledger. It's a big thesis.
But it was abstract. When I read it, I thought: if this is true, where's the data? Where is the moment where internet-native financial infrastructure did something that the existing system literally could not?
That moment happened on February 28th, 2026 at 6:14 AM UTC.
Saturday, 06:14
The United States began striking Iran at approximately 6:14 AM UTC on a Saturday. That's 1:14 AM on the East Coast. The CME had been closed for eight hours. The NYSE wouldn't open for another 46 hours. The most sophisticated capital markets infrastructure ever built: trillion-dollar exchanges, microsecond matching engines, armies of quants, had absolutely nothing to say about the start of a war.
But the internet was open.
Five different venues were actively trading crude oil, gold, equities, and geopolitical risk on crypto infrastructure. Perpetual futures on Hyperliquid. Weekend CFDs on IG. And prediction market contracts on Polymarket, where $529 million had already been traded on the timing of US strikes on Iran. The media narrative in the days that followed focused almost entirely on Polymarket. Six freshly created wallets had bought "yes" shares at 10 cents hours before the bombs fell, making over $1 million in profit. The story was about insider trading, about the ethics of betting on war, about whether prediction markets should be regulated out of existence. What nobody asked was the market structure question: who actually priced the event first?
We pulled sub-second order book data from Hyperliquid and Binance, minute-level data from Polymarket, and reconstructed the full sequence. Here's what happened.
06:13:59: Everything is dead. BTC on Binance: $65,580, trading 2-3 trades per second. CL crude oil on Hyperliquid: $68.71, flat. Polymarket "Strikes by Feb 28": 21.5%. It's Saturday night. The world is asleep.
06:14:34: BTC information arrival. 607 trades in one second. 5.4 BTC of volume. Price crashes $54 to $65,526. This is unmistakable: this is where the most liquid crypto market on earth learned about the strikes.
06:15:41: CL's real move. The bid jumps from $68.73 to $68.99. Both sides of the book move simultaneously, this isn't one ask getting lifted, this is the entire market repricing. Mid hits $69.04, up 0.48%.
06:16:25: CL breaks +1.05%. Less than two minutes after BTC's first dump. A decentralized exchange with a fraction of CME's volume has priced in a major geopolitical event. On a Saturday. With no reference market.
06:18:00: Polymarket confirms. "Strikes by Feb 28" jumps from 21.5% to 81.5%. CL moved first. BTC moved second, 33 seconds later. Polymarket confirmed four minutes after that. The prediction market, the one with $529 million of volume, the one that dominated every headline for a week, was last. It didn't predict the strikes. It confirmed them, after the asset markets had already repriced. The real Polymarket move came at 06:18, by which point CL was already up over 1%.
Not one weekend. Every weekend.
The Iran strike is a dramatic case study. But the structural claim, that internet-native markets are becoming the price discovery layer when traditional markets close, requires more than one data point.
On the weekend of March 8-9, we ran a comprehensive comparison across every venue that trades crude oil when CME is closed: xyz:CL on Hyperliquid, OKX's CL-USDT perp, Binance's (Aster) CLUSDT perp, and IG's weekend crude oil CFD. True L2 order book data on the crypto venues. Minute-level on IG.
The three crypto perp venues - Hyperliquid, OKX, and Binance - are effectively one market. The basis between them averages under 20 basis points. Return correlations above 0.6 at the one-minute level. They move together, they price the same information, and they do it continuously.
But they don't move simultaneously. Using L2 mid-to-mid lead-lag analysis, xyz:CL on Hyperliquid leads OKX at every timescale we tested: At 1-second lag: CL leads 1.1x. At 5 seconds: 1.4x. At 10 seconds: 1.8x.
A decentralized perpetual exchange deployed by a team that staked 500,000 HYPE tokens, with no listing committee, no regulatory approval process, no exchange membership, is the price discovery venue for weekend crude oil. OKX, a centralized exchange with billions of dollars of daily volume, follows it. This is the thesis about permissionless innovation made quantitative.
IG: the control group
IG is a publicly traded, FCA-regulated company with a market cap of several billion dollars. They offer weekend trading on crude oil, marketed as a way to manage risk when traditional markets are closed. Their clients include retail traders, wealth managers, and small institutions across Europe and Asia.
IG's weekend crude oil CFD has a return correlation with Hyperliquid's CL perp of 0.06.
That's effectively zero. At the one-minute level, IG's weekend product is completely disconnected from real-time price discovery. It trades at a systematic discount of 147 basis points to the crypto perps, with hour-by-hour variation ranging from -19 to -391 bps. The median spread is 64 basis points, 23 times wider than Hyperliquid's 2.8 bps.
At the 30-minute level, a weak positive correlation emerges (r = 0.26). IG follows the same broad direction, eventually. But the lag is enormous.
When CME reopened on Sunday at 22:00 UTC, the dislocation was laid bare. Anyone using IG's weekend product as their primary view on crude oil was working with a price that was 7% wrong.
This is not a criticism of IG as a company. It's a structural observation. IG's weekend product is a synthetic market, a model-derived price, not a price that emerges from supply and demand. The crypto perps, for all their limitations in liquidity and infrastructure, produce prices that are formed by actual trading, actual order flow, and actual information. The difference in quality is measurable and it's large.
The infrastructure gap
If Internet Capital Markets are already winning at price discovery, why isn't the transition further along? Because the infrastructure isn't complete yet.
The perps work. The prices are accurate. But they're derivatives, they need a spot market to anchor to. During the week, that anchor is CME and NYSE. On weekends, it doesn't exist. The oracles that feed these markets go stale, the mark prices drift, and the risk management layer degrades even while the price discovery layer performs.
The incumbents know this gap is real. NYSE and Nasdaq have both announced plans to move toward 24/7 trading. That's not a coincidence: it's a direct response to the reality that price discovery doesn't stop on Friday afternoon anymore, and the exchanges that close are losing relevance to the ones that don't. When the NYSE eventually trades around the clock, the weekend oracle problem disappears overnight. The perps get a live reference, the infrastructure gap closes, and Internet Capital Markets go from "surprisingly good on weekends" to "indistinguishable from traditional markets, but permissionless."
Tokenized spot assets, stocks, commodities, ETFs living on-chain with real redemption mechanisms will likely get there even faster. The infrastructure to trade a tokenized S&P 500 around the clock already exists. The regulatory framework is catching up. Either path leads to the same destination: a world where the concept of "market hours" is an artifact of 20th century infrastructure, not a feature of 21st century capital markets.
What this means
Three years ago, if you'd told a crude oil trader that a decentralized exchange built on a blockchain would lead price discovery on weekeds, and make IG's weekend product look like a toy, they'd have laughed at you.
The data says otherwise.
When traditional markets close, Internet Capital Markets are already the most efficient price discovery mechanism available. Not in theory. Not in a pitch deck. In the data, at sub-second resolution, on a Saturday morning when the world is asleep and a war has just begun.
The Internet Financial System isn't coming. It's here. And it's open on weekends.
Data sourced from https://pragma.build (Hyperliquid/Polymarket L2), CoinAPI (OKX and Binance L2 BBO), Aster API, IG API,
@hydromancerxyz. Iran strike analysis covers February 28, 2026. Lead-lag and venue comparison covers March 8-9, 2026. All errors are mine.