Treasury said it will at least double its long-dated buybacks from September and the long end rallied, which was the loudest number of the day. The quieter pattern was about electricity. A bitcoin miner's real asset turned out to be its grid connection. A chipmaker guided high on the parts between the racks. A governor made a hundred-megawatt data centre subject to a community vote. The scarce input in artificial intelligence has moved from the chip to the delivered watt. Most of the day owes that pattern nothing: a cancer result, a mortgage book, one changed word in Malaysia.
A three-day losing streak ended on a health-care melt-up with technology leading the decliners, an index rising on the sector nobody owns for growth. The dollar gave up 0.86 percent in the same session the long end got its relief, and taking lower yields while selling the currency reprices who ends up holding the duration. Crude stocks built a third straight week, to 428.8 million barrels. Bitcoin touched $69,000 and ether reclaimed $2,000 while $1.92 billion of leverage was closed out around them, nine tenths inside four hours, which is positioning rather than conviction.
More than thirty AI models now sit above the European AI Act's compute threshold, which obliges their makers to notify Brussels within two weeks. Call the mechanism verification exposure: the access that produces assurance is the same access that produces intelligence, so the monitored party prices an inspection as a disclosure. A nuclear test ban died in 1963 over three on-site inspections a year against seven, and what got signed excluded the only category that required letting anyone in. We expect AI governance to split along access rather than ambition: declaration-based regimes get agreed quickly, anything requiring standing entry to weights or a facility stalls. The case against us is the Chemical Weapons Convention, which designed bounded access into the instrument and got signed anyway. An honest engineer and an unwilling party produce the identical sentence when they say a scheme cannot be verified.
Give E. coli all the oxygen and sugar it wants and past a certain growth rate it stops burning sugar all the way down, fermenting it instead and throwing most of the energy away. Markus Basan and colleagues measured why in Nature: respiration yields far more energy per molecule but takes twice as much protein to run, and protein is the fixed budget. The cell is careful with apparatus, not careless with food, and which resource to conserve depends on which is scarce.
Nobody has bridged the gap to the far smaller count used to run an algorithm. Holding a thing and getting work out of it are separate achievements.
Arizona State researchers mapped 69,000 species onto two phylogenies and found nearly the whole signal comes from one clade of wasps and ants. It survived on elegance rather than evidence.
“A beginner should not hold two arrows. It will make him rely on the second arrow and be careless with the first.”
– Yoshida Kenkō, Essays in Idleness, c. 1330
You would call the archer prudent, not lazy. He brought a spare because things go wrong. His teacher says the spare is the problem.
The claim is stranger than it looks. It is not that the second arrow makes him fail. It is that holding it changes what his hands do during the first shot, before anything has gone wrong, in a way he cannot feel.
You are holding one right now, and you know which. The second job application open in another tab. The version of the conversation where you do not have to say the hard sentence. The cost is not paid when you fall back on it, but continuously, by the part of you that stays behind to keep it warm.
Today's practice: find the second arrow in the work you are making right now, the safer version you built in case the real one fails. Remove it before you next sit down.
In 1693 William Molyneux asked John Locke whether a man born blind, who had learned by touch to tell a cube from a sphere, could tell them apart by sight alone once he could see. Philosophers argued for three centuries. Then a group at MIT tested newly sighted patients days after surgery: near chance at first, right within five.
Some questions are like that: the method was fine and an observation was missing. Others differ in kind. Write the best answer your method can give, in its own vocabulary, and see whether the question still stands. If it does, you are short of vocabulary, not data.
The decision tool: run that test before you spend. If it collapses, go get the observation. If it survives, measuring harder will feel like progress and will not be.
Explore this model →Most of what mattered today happened in layers nobody looks at. Worth remembering next time a headline looks obvious. See you tomorrow.