California has designed a border that cannot be crossed anymore. It is January 1, 2026.

Proposition 40, now headed for the November ballot, would impose a one-time tax equal to five percent of the net worth of billionaires who were California residents on that date. The bill would come due in 2027. A taxpayer could stretch payment across five years, at additional cost, but moving to Nevada in February would not erase the obligation. The political campaign is happening in the future tense; the proposed liability has already chosen its past.

That is the measure’s most consequential feature. California is not merely asking voters whether billionaires should pay more. It is trying to prevent the familiar pre-tax choreography in which a fortune relocates just before the music stops.

The tax would generally cover businesses, securities, art, collectibles and intellectual property while excluding real estate, pensions and retirement accounts. Ninety percent of the proceeds would be reserved for health care, with the remainder going to education, food assistance and administration. Its labor backers have promoted a $100 billion goal, chiefly to reinforce Medi-Cal after federal cuts. The nonpartisan Legislative Analyst’s Office is more cautious: it expects tens of billions spread over several years and warns that subsequent departures or other tax planning could reduce annual income-tax collections by less than $1 billion.

This is a spectacular wager on arithmetic. California depends on the top one percent for nearly half its personal income-tax revenue. The initiative says the state can take an extraordinary slice of accumulated wealth without fatally damaging that recurring stream. Opponents say the slice will teach the stream to flow elsewhere.

The people with the greatest exposure have not waited for November. Google co-founder Sergey Brin contributed $82 million to Building a Better California, a committee supporting measures intended to blunt the tax. The committee had raised more than $118 million from fewer than a dozen donors, according to the Associated Press. That concentration is itself a campaign advertisement for the tax: a tiny electorate of fortunes can finance a political apparatus before most voters have read the ballot label.

Governor Gavin Newsom opposes the proposal, as do organizations including the California Medical Association and California School Boards Association. Their argument is not simply that rich people dislike taxes. It is that California would exchange a durable, if volatile, income-tax base for a one-time collection entangled in valuation fights and litigation. The union behind the initiative offered a two-percent compromise; Newsom still declined.

Both sides therefore agree on the uncomfortable fact beneath the slogans. California’s public services are unusually dependent on a small population with unusually portable wealth. The dispute is over whether that dependence should command deference or justify collection.

Then there is the valuation problem. A public share has a price at the closing bell. A private company, a controlling interest, an intellectual-property portfolio or a collection of art has an argument. At this scale, every appraisal becomes a lawsuit with a decimal point. The state estimates administration could cost tens of millions annually for several years. That is not evidence the tax cannot work. It is evidence that “five percent” is the beginning of the calculation, not the end.

Proposition 40 has made time part of tax policy. January 1 determines who is inside. December 31 helps determine what the property is worth. Election Day decides whether the obligation exists. And 2027 is when the government asks for the money.

A fortune may be mobile. A calendar is not. California is betting that voters will prefer the authority of the date to the velocity of the money.

President Donald Trump is trying to describe an exit from the Iran war as a collection.

Iran has sought compensation in peace discussions. Trump answered that if Tehran wants damages, the United States will demand them too. The symmetry is rhetorically tidy and materially false. Negotiators are not dividing an insurance settlement. They are bargaining while American forces enforce a blockade, oil remains above its prewar price and Washington consumes the stockpile intended to soften genuine supply emergencies.

The administration says nearly nine million barrels a day are again moving through the Strait of Hormuz, with regional flows reaching roughly 15 million when pipelines are included. Yet American oil futures were still around $83 a barrel in the latest figures reported by the Associated Press. The Strategic Petroleum Reserve had fallen below 300 million barrels—more than 100 million fewer than at the start of the year.

At sea, the coercion continues. U.S. Central Command said American forces fired into the engine room of the Panamanian-flagged *Vela Nova* after it ignored orders to stop in the Gulf of Oman. The military says it has redirected 55 commercial vessels, disabled three and boarded two while more than 15 American warships operate in the Arabian Sea.

This is not yet the posture of two governments pricing a clean departure. It is a negotiation conducted through ships, insurance premiums and a reserve tank whose level can be measured every Friday.

Iran can prolong the argument by making passage uncertain. Washington can prolong it by spending military power and stored oil. Neither resource is infinite. The decisive peace term may not be the compensation figure announced at a lectern, but the inventory each side can still afford to expend before signing.

For two decades, Chuck Redd presided over holiday jazz performances at the Kennedy Center. After President Donald Trump’s name was added to the building, the drummer and vibraphonist canceled his Christmas Eve appearance in protest. The performing-arts venue sued him.

The lawsuit failed. Now the invoice has arrived.

D.C. Superior Court Judge Tanya Jones Bosier, who dismissed the Kennedy Center’s case in June, ordered the institution to pay approximately $252,000 in Redd’s legal fees and costs within 45 days. Redd’s attorney called the original suit a meritless effort to silence opposition. The Kennedy Center says it will appeal.

The order matters beyond one musician’s fees because political capture often works through transaction costs. An institution does not need to win a dubious case if defending it is expensive enough to frighten the next performer into compliance. Fee awards reverse that pressure: the organization that selected litigation as an instrument must bear the price of the instrument.

Redd lost a booking and acquired a legal fight. The Kennedy Center sought to turn a cancellation into discipline. For now, the court has turned it into accounts payable.

The Strategic Petroleum Reserve is commonly discussed as if it were a presidential verb: release, refill, stabilize. It is actually a finite inventory in salt caverns.

That inventory fell below 300 million barrels last week, according to figures reported by the Associated Press—down more than 100 million since the beginning of 2026. The draw has helped cushion oil markets during the Iran war and the disruption around the Strait of Hormuz. It has also consumed roughly a quarter of the reserve’s opening stock in a little over seven months.

The administration points to improving flows. Energy Secretary Chris Wright said close to nine million barrels a day were moving through Hormuz and that total regional flows, including pipelines, averaged about 15 million. But U.S. crude remained around $83 a barrel, above its prewar level.

The reserve does not have to run dry to lose power. Its deterrent value declines as traders calculate how many additional disruptions Washington can cover and for how long. Each release buys current calm by selling future optionality.

That trade may be justified during a war affecting the world’s most important oil passage. It is nevertheless a trade, not a free demonstration of control. Replacement barrels could cost more than the barrels sold. Cavern operations have physical limits. Another hurricane, refinery outage or geopolitical rupture would arrive at a government already using its emergency cushion for the present emergency.

Markets are often accused of ignoring politics. Here they are reading it precisely. The price of oil incorporates not only the number of tankers passing today but the shrinking quantity Washington can release tomorrow. The reserve has bought time. The balance sheet shows how much time was burned.

Twitch has added a privacy control with an unusually revealing purpose: creators may now tell Amazon not to use their work to train generative-AI systems.

The default is permission.

Twitch’s own account materials say eligible content can include streams, video on demand, clips, chat, images and text. Training may improve tools used on Twitch, such as transcription, or models deployed elsewhere across Amazon. Switching the control off prevents future use for generative-model training, but it does not end every machine-learning use on the platform.

Chief product officer Mike Minton supplied the corporate logic with a clarity most privacy policies avoid. If the program required affirmative consent, he said during a livestream, nobody would opt in.

That sentence collapses the distinction between consent and capture. The company knows the preference of the people supplying the material, then designs the interface to obtain the opposite result from those who do not discover the setting.

The rollout was disclosed indirectly through an announcement that an opt-out had been added, rather than through a prominent declaration that creator material was being used. Twitch also attempted a corporate distinction: Amazon, not Twitch, trains the generative models. Creators are unlikely to find much comfort in the internal routing of data inside the company that owns their platform.

The commercial attraction is plain. Livestreams contain vast quantities of conversational speech, improvised performance, audience reaction, moderation decisions and parasocial interaction. They are not simply videos. They are labeled rehearsals for systems meant to speak, entertain and hold attention.

Twitch creators already provide the show, the audience and a share of the platform’s revenue. The new setting reveals a fourth product being extracted from the same performance: training material.

The useful question for every platform is no longer whether an opt-out exists. It is why the company needed the default to contradict the preference it openly admits its users hold.

A normal book buyer leaves a pattern. Gardening. Naval history. First editions. A particular author whose lesser novels must be defended at dinner.

The new buyers haunting secondhand shops leave inventory lists without taste.

Booksellers in Britain and Ireland have reported large orders combining titles with no visible relationship: an Estonian translation of John le Carré’s *The Mission Song*, a particular edition of Anne Brontë’s *Agnes Grey*, an October 1983 issue of *Warship*, books on agricultural equipment in 18th-century Africa and biographies of 1950s racing drivers. One unidentified dealer said related buyers had ordered 6,000 books since January, paid full price and sometimes used different names for deliveries to the same address near Heathrow.

No purchaser has publicly confirmed the theory moving through the trade: that AI companies or their contractors are buying books to scan and pulp them. The suspicion is still suspicion. But it has a history. Anthropic has acknowledged that sourcing books is a widely used method of assembling language-model data and says its programs do not buy and destroy rare or antiquarian volumes.

The oddness of the lists is the clue. A human library expresses an appetite. A machine corpus expresses a coverage gap.

Books published before the generative-AI boom possess a new commercial virtue: they are unlikely to contain generative-AI prose. Obscure printed works that once looked economically inert may now be clean samples, unavailable in searchable digital form and useful precisely because no algorithm has previously digested them.

This gives the secondhand shop a strange final role in the information economy. It is not merely a picturesque refuge from the screen. It is a warehouse of unmodeled language.

The buyers may be paying well, and some dealers understandably welcome the orders. Yet the possible conversion is brutal: a book that survived decades of owners is purchased as raw material, its spine removed so its contents can be made statistically immortal while the object itself is destroyed.

The machine does not collect books. It eliminates the reason to return them.

Dario Amodei has offered the best sentence an AI executive has produced this year: the industry will not regain trust by saying it can cure cancer, but by actually curing it.

He also offered himself an extremely generous deadline.

The Anthropic chief argues that public hostility toward AI reflects a broader crisis of trust and concedes that companies have not delivered their largest promised benefits. He says Anthropic is accelerating work in biology and medicine, with early indications possible in the coming months and larger results later.

The candor is welcome. The framing is not innocent.

“Cure cancer” transfers judgment from the present, where AI companies are building power plants, collecting copyrighted material, automating work and lobbying governments, to a hypothetical future breakthrough so morally magnificent that today’s conduct may look petty by comparison. It turns accountability into a clinical-results waiting room.

A cancer therapy would deserve celebration. It would not retroactively answer who paid for the electricity, whose data trained the system, which patients could afford the treatment or whether the same company distorted labor and information markets along the way. Social permission is not a medal awarded to whichever industry produces the grandest invention.

Amodei is right that better marketing cannot repair distrust. But trust is not withheld merely because the miracle is late. It is withheld because ordinary people correctly observe that companies demand broad authority before producing narrow evidence.

The proper standard is smaller and harder: show the work, publish the limits, compensate the suppliers, share the gains, accept liability and permit refusal. Do that while pursuing the cure.

Humanity should not have to choose between medical progress and enforceable terms of service. Any machine intelligent enough to transform biology ought to be operated by institutions intelligent enough to understand that distinction.

The United States has removed more than 100 million barrels from the Strategic Petroleum Reserve this year while fighting Iran and policing the Strait of Hormuz. Officials present the releases as proof that the government can contain the economic damage.

They prove the opposite: the damage is being transferred into the future.

Emergency stocks exist so a temporary rupture does not become an immediate national crisis. They are not a substitute for an achievable war aim. If military policy requires a continuing public sale of stored oil merely to keep the price politically tolerable, then the oil market is supplying the cost estimate that the administration will not.

The reserve is now below 300 million barrels. The significant number is not zero. It is the point at which every new release makes the next emergency harder to insure and every promised refill more expensive to execute.

Congress should require a replenishment schedule and explicit war-linked accounting for each further draw. Voters deserve to see military expenditures and reserve depletion on the same page. One is paid in dollars now; the other is paid in diminished choices later.

A government may reasonably use emergency oil during a war. It may not honestly call the resulting calm evidence that the war is sustainable. The gauge is part of the strategy, whether the White House reads it aloud or not.