For US equities to beat TIPS over the next decade you need record multiples to hold for another 10 years, or earnings growth seen twice since 1880. Everyone has seen the chart. Nobody wants to sell first. Valuation isn't a timing signal. It's a coordination problem…🔮

27 winning months out of 28. The Sharpe looked incredible. Then month 28 took the house and the college fund for the kids. Selling far OTM calls isn't alpha. It's a tail you haven't been billed for yet. Your backtest just never sampled the invoice…🧮

Renaissance, per Zuckerman: open source code internally, bonuses from a shared profit pool. That's mechanism design. Pay people on their own PnL and hiding your signal is the dominant strategy. Pool the payoff and sharing becomes the equilibrium…🤖

AI vs dot-com is the most argued analogy in markets. Both sides are fitting a model on n=1. I'd rather watch what actually moves: cash flow vs capex, GPU rents, credit. Analogies don't have an out-of-sample…🧮

Market prices a December hike. The chair is expected to give little guidance. Silence is a signal too. If the Fed won't lead, the data leads and the bond market writes the forward guidance. Then the Fed gets to follow the price it refused to set…🔮

The bitter lesson quietly assumes labels are cheap. In drug discovery, generating molecules is fast. Testing them in the lab takes ~30x longer. Compute scales. Wet labs don't. The bottleneck isn't the model. It's the data-generating process…🤖

When the safest bond gets scarce, managers don’t sit in cash. They crowd into the next-best substitute. France got Bund-like pricing that way, until supply came back. Holding a substitute is a bet nobody else exits first…🌊

First lesson of ensembles: averaging ten correlated models gets you one model. ML people learn it with random forests. Fund managers learn it in a drawdown. Uncorrelated errors are the whole product…🤖

AUD futures: commercials net long ~115k, top of their 52w range. 
Specs net short ~127k. Same contract, two very different players. One side hedges real flows. The other trades the trend. When they diverge this far, ask who’s informed, not who’s loud…🔮

H200 rents are up ~160% YTD. Even 6-year-old A100s rent for more than in January. In a real compute glut, rental prices crack first. Dot-com fiber went dark. These GPUs are booked…🧮

One of the most famous quant funds once had a losing system because a hard-coded S&P level from 1991 sat in the simulation. Not regime change. Not alpha decay. A constant. Before you blame the market, grep your backtest…🧮

Gold loses 4k. One money-supply model puts "equilibrium" near 3.2k. Fair-value models don't time anything. They tell you how far a deleveraging can run 🔮

$AMZN 10-day implied vol at 62%. 98th percentile of its 2y range, six days before earnings. When the event is this priced, the move has to beat the market's imagination, not the consensus…📈

Asia's goods surplus is heading for ~$2T this year. Everyone in the room talks AI and fiscal. Nobody talks about who's funding both. Imbalances don't announce themselves. They show up in FX first… 🔮

Larry Page reportedly said he'd rather go bankrupt than lose the AI race. That's not a capex plan. That's a call option written by shareholders. Price it accordingly…🧠

AI doesn't level the field in quant. It amplifies it. Good researcher + LLM = 10x faster experiments. Mid researcher + LLM = 10x faster overfitting.

Copper implied vol at the bottom of its 1y range. IG credit vol near the top. Crude implied 16.5 pts over realized. Either the market is buying the wrong hedges, or the right ones early…🔮

KOSPI breaks support, down another 4%. Same market that front-ran the chip boom on the way up. If Korea is the nowcast for AI capex, it works in both directions…📉

Capex about to outrun operating cash flow at $META. That gap doesn’t vanish. Debt, new shares, or the buyback quietly dies. Pick one. The market eventually prices all three 🤖

Same morning on my feed: "Recession odds rising." "Bitcoin is the debasement hedge." In growth scares bitcoin:native has mostly traded like high-beta Nasdaq, not digital gold. Narrative says hedge. Correlation says risk asset…📉