Bounced, Not Cleared

“The things you cannot measure are not the things that do not matter. Usually they are the ones that matter most.”

Equities bought the dip from the post-FOMC selloff while the bond market quietly flattened, two time horizons reading the same data and reaching opposite conclusions. Ukraine launched its largest drone attack on Moscow since the full-scale invasion, Noam Shazeer left Google for OpenAI, and the hottest retail and manufacturing data in months arrived into a central bank that just stopped telling markets what comes next.

Checking for audio...

S&P

—

NDX

—

DOW

—

BTC

—

ETH

—

SOL

—

Gold

—

Oil

—

10Y

—

Markets minute

US markets are closed today for Juneteenth and reopen Monday, so these levels reflect Thursday's close. Equities bought the post-FOMC dip but the Russell lagged, marking a positioning unwind, not returning conviction. Bitcoin held near 64,000 while stocks jumped over a percent, snapping two weeks of tight correlation. The ten-year held 4.49% and the curve flattened from the long end, a shape that prices a policy mistake, not a soft landing. When equities and bonds read the same data and reach opposite verdicts, trust the side with the longer memory.

Today's signals

Stocks Bounced. The Bond Market Refused to Sign Off. The S&P recovered 1.08% to 7,500 and the Nasdaq jumped nearly 2%, a clean post-FOMC dip-buy. But the bounce came on no new information, and the bond market did not follow. The curve flattened from the long end, the shape that preceded the policy reversals of 1995, 2006, and 2019, not the soft landings markets like to remember. Michael Howell calls this the Speculation phase: prices climbing on momentum after the liquidity cycle has already turned, historically the last leg before a correction. The 2018 tape is the precise warning. The S&P bounced 4.9% in two sessions after Powell turned hawkish that October, then fell 20% by Christmas. Two time horizons are reading the same data and reaching opposite verdicts, equities seeing a buyable dip, bonds seeing a mistake forming. The resolution is mechanical. If the S&P prints a new high by July while the curve keeps flattening, the bounce was real. If the ten-year breaks 4.60% while stocks stall, this was the last rally before the catch-down.

geopolitics

AI's Bottleneck Isn't Chips Anymore. It's the Neighbors. The consensus says the limit on AI is chips, then power. Both are largely solved: silicon is allocated, roughly $725 billion is committed for 2026, and power is contracted. The real bottleneck is whether a county board says yes. Capital raised in a weekend now sits stranded for years behind a public comment period. Operators are mispricing that consent by orders of magnitude because they file it under PR and fund it like PR. The arithmetic is absurd. A 1.6 gigawatt datacenter throws off about $3 billion a year, so paying every resident of its host town $10,000 annually costs under 4% of revenue. Yet OpenAI's Stargate Michigan put roughly $10 million on a $50 billion project, and the marquee perk was a rec-center lazy river. Expect a top-five hyperscaler to reframe a build around per-resident revenue-sharing by Q4, and watch its permitting clock beat rivals still treating consent as charity. Where it breaks: if the real fight is water tables and trust rather than cash, a check reads as a bribe and hardens the no.

crypto · defi

Ukraine Moved the War Into Moscow's Gas Tank. Ukraine launched its largest drone attack on Moscow since the invasion began, sending more than 550 drones at the Kapotnya refinery and setting a key fuel site ablaze. Kapotnya supplies over a third of Moscow's fuel, so this was not symbolism. It was logistics with civilian consequences inside the capital. Russia claims it intercepted 555 drones nationwide and 194 over Moscow alone, a volume that overwhelmed layered air defense and signals Ukraine has solved mass-simultaneous launch at production scale. The strategic shift is the story. Ukraine has moved from military and border targets to the economic assets that fund the war, and placed them inside the city where the decisions get made. When airports suspend flights and apartment buildings report damage, the war stops being something Russian civilians experience only through the news. If Ukraine sustains refinery strikes at this scale through summer, expect domestic fuel prices to climb and the political calculus to shift, slowly, from the Kremlin toward the population that now feels it.

crypto · defi

Four Identical Charts, Four Completely Different Promises. Solana has won the venue for tokenized equities with roughly 97% of spot volume, but the four dominant issuers wear identical price charts over radically different legal instruments, and the difference stays invisible until something breaks. Backpack's SPCX redeems into the real share through DTCC plumbing. Ondo issues exposure as a structured note. Backed Finance's xStocks is a bearer tracker carrying issuer credit risk. PreStocks was a synthetic promise backed by special-purpose vehicles. In May, Anthropic and OpenAI voided those vehicles, the tokens fell 34 to 40%, and the platform had been showing an implied $1.3 trillion Anthropic valuation against $23 million of real assets. The Lehman echo is exact. Its principal-protected notes in 2008 carried the same architecture: holders thought they owned the index, but they owned an unsecured promise, and when the issuer failed the wrapper was worthless while the underlying was fine. Tokenization does not abolish counterparty risk. It relocates it into the wrapper, where the chart looks identical right up until it doesn't.

crypto · defi

Open-Source AI Just Pulled Within a Point of the Frontier. Z.ai released GLM-5.2 under an MIT license with open weights, and the scores are the story. It posted 62.1 on SWE-bench Pro, beating GPT-5.5, and landed within a single point of Claude Opus 4.8 on the agentic coding benchmark, all at roughly one-sixth the API cost. The lag from frontier model to open-weights equivalent has compressed from quarters to weeks. The structural consequence is pricing power. If two or three more open-weights models reach the top five on agentic benchmarks this quarter, the premium closed providers charge erodes from a moat to a convenience fee. That is the commodity cycle: once supply turns abundant, the input stops commanding a margin, and enterprise budgets shift from buying model access to building deployment infrastructure. Inference providers are already listing GLM-5.2 at prices that undercut closed-model calls by an order of magnitude. The labs raising tens of billions on model supremacy are now racing a clock that open weights keep resetting.

ai · tech

Interesting things

Fourteen Sloppy Shuffles Beat Seven Perfect Ones.

A new proof shows that about 14 of the messy, uneven riffle shuffles real people actually do are enough to fully randomize a deck, updating the famous 1992 result that required seven mathematically perfect ones. The deeper find is about how order dies. The deck stays mostly sorted through the first dozen shuffles, then tips into full randomness in the final few, a sharp phase transition rather than a gradual fade. Disorder does not creep in. It arrives all at once.

Scientists Built a Material That Refuses to Choose.

Strength and flexibility almost always trade off against each other, but researchers found that tiny U-shaped, staple-like particles tangle into a material that is both at once. They lock through geometric entanglement rather than chemical bonds, the same trick that makes chainmail rigid under impact yet fluid in motion, which means it can be taken apart and rebuilt without degrading. If it scales, it points to recyclable structural materials that skip the energy-hungry bonding metals and composites demand.