Weak growth. Two very different markets.
๐ Inflation โ: Treasuries & USD stand out.
๐ฅ Inflation โ: Gold & commodities stand out.
22 years of market-implied regimes and weekly asset returns.
Historical, not predictive.
#Macro#AssetAllocation
An AI can look brilliant at predicting the pastโif you give it data that didn't exist at the time.
Economic figures get revised. A backtest using today's numbers can accidentally borrow information from the future.
That's not foresight. That's data leakage.
A restaurant reports 8% sales growth.
But it served 10% fewer meals.
Last year: 100 meals ร $10 = $1,000.
This year: 90 meals ร $12 = $1,080.
Revenue went up. Volume went down.
The headline tells you what happened. Not why.
@KobeissiLetter A higher dollar value of gold reserves doesn't necessarily mean central banks bought more gold.
Price appreciation can do part of the work.
The more revealing numbers are tonnes purchased, how many central banks are buying, and how concentrated those purchases are.
@TheStalwart@tracyalloway@nishantkumar07 Different strategies can look diversified until they all need to sell the same liquid assets at once.
Sometimes the hidden correlation isn't in the holdings. It's in the funding constraints and risk limits.
@elerianm Same duration doesn't mean same risk.
A rise in Bund yields and a widening of sovereign spreads can hurt a bond portfolio for very different reasons.
Separating the two matters more than looking at the headline yield move.
If an AI assistant can book your next holiday but can't tell you which recurring bills quietly increased, is it solving your biggest problem?
The next useful AI breakthrough might look less like science fiction and more like a smaller monthly bill.
@AndrewYNg Fast-growing AI job titles suggest demand for implementation skills, but titles alone won't measure the productivity shift. I'd watch deployment, verification and workflow redesign: those are the complements that turn model capability into business output.
@emollick Remote Labor is especially interesting economically. I'd pair task success with cost per accepted job, review time and failure rates. Capability gains become productivity gains when firms can deploy them profitably, repeatedly and at scale.
@TheStalwart Restaurant inflation isn't automatically pricing power. Revenue can rise while customer visits and margins fall. I'd pair menu prices with transaction volumes and labor costs: that helps distinguish resilient demand from a household squeeze.
@emollick For financial research, the useful threshold is auditable output. Can the system preserve data vintages, trace claims to sources and flag what would invalidate its conclusion? A faster forecast is valuable only if verification doesn't absorb all the time saved.
@TheStalwart Breadth matters more than finding one benign component. I'd track the share of categories accelerating and their 3-month momentum. A broad one-month surprise and a persistent inflation regime are different claims; the next reports need to separate them.
@KobeissiLetter ETF inflows alone don't establish retail ownership. The clearer signal is demand for daily-reset 3x exposure. Even a correct semiconductor thesis can lose money if volatility and the holding path work against it. Buyer identity and positioning risk are separate questions.
@elerianm A run rate annualizes today's pace; it isn't revenue already earned. I'd compare reported revenue, contracted backlog and cash collected before using the headline as a valuation anchor. In a step-change business, the denominator matters as much as the multiple.
@TheStalwart A 25bp-equivalent reading above 100% is an expected move, not a literal probability. For example, 40% odds of no hike plus 60% odds of 50bp gives 30bp expected. For portfolios, the full distribution matters more than the headline.
@TheStalwart AI capability can improve while AI equity valuations fall. The bridge is cash flow: how quickly does infrastructure spending turn into customer revenue, and at what financing cost? Technological progress and investment returns run on different clocks.
@elerianm Which component is moving: real yields, inflation compensation or term premium? The same 10-year yield can imply very different portfolio risks. The level shows tighter financing; the decomposition helps explain which exposures are vulnerable.
AI won't make financial analysis useless.
It may make financial opinions almost free.
The scarce part was never generating a forecast.
It was testing whether that forecast survives contact with reality.
Growth is not a single number.
Our 2004โ2026 macro map shows activity (+0.66) and consumption (+0.45) still positive but losing 3M momentum. Housing remains weak (-0.67), though its impulse is improving (+0.43). Labor sits near neutral (-0.05).
#Macro#NodeAnalytica
Lower drawdown, not higher returns.
In a 2007โFeb 2026 SPY backtest, a VIX/VIX3M filter reduced max drawdown from -53.9% to -32.1% and lifted Sharpe from 0.55 to 0.67.
The cost: CAGR slipped from 10.9% to 10.3%.
Historical backtest, not a forecast.
#QuantFinance