The most consequential word in New Mexico’s case against Facebook may be alone.

Nearly a decade after the Cambridge Analytica scandal exposed how information gathered through Facebook applications could be carried into political targeting, New Mexico has pushed its lawsuit against Meta toward a four-week trial. Other governments investigated, negotiated and moved on. This one has brought jurors into the room.

The state’s case concerns Facebook’s handling of third-party applications and the personal information those applications obtained. Meta argues that it investigated suspicious apps after the breach and that New Mexico is presenting an obsolete picture of a platform whose safeguards have since changed. That defense is more substantial than a denial: it asks the jury to treat subsequent repairs as evidence about the condition of the machine when the damage occurred.

This is why a trial matters. Settlements can compensate, constrain and conclude. They rarely reconstruct. A trial makes competing accounts occupy the same calendar. Executives, investigators and technical systems become parts of one evidentiary record rather than separate rounds of statements issued years apart.

New Mexico’s persistence is especially notable because the machinery of accountability has been shrinking around it. The scandal produced hearings, fines, consent orders and a vast public vocabulary about data privacy. It also produced fatigue. People learned that their intimate preferences could be inferred, packaged and sold, then continued using the products because social infrastructure does not become optional merely because its business model is alarming.

Meta’s argument that its protections improved deserves examination. Platforms do change. Controls added after a scandal can reduce the likelihood of repetition. But improvement and responsibility are not mutually exclusive findings. A fire code written after a blaze does not answer who left the door locked.

The trial also presents a problem larger than Facebook. Technology regulation increasingly arrives through isolated prosecutors, private plaintiffs and judges because legislatures and national regulators move too slowly for systems that can be redesigned between hearings. That arrangement produces occasional breakthroughs, but it is a poor substitute for durable rules. One determined state should not have to carry the public history of an industry.

Still, there is value in the lonely case. Institutions often display their actual priorities not when a scandal erupts, but when the television crews depart and the procedural costs begin. New Mexico has chosen discovery, testimony and the risk of losing in public. The outcome remains for the court and jury. The achievement so far is simpler: somebody declined the convenient ending.

Nepal observed a national day of mourning after the August 26 floods killed more than 1,300 people in Nepal and Tibet. At the time, roughly 5,000 remained missing, according to Nepal’s disaster-management agency. The scale turns every official number into a provisional one.

Floodwaters swallowed settlements and severed mountain roads and bridges. The physical destruction is also a political examination for Prime Minister Balen Shah, 36, who took office in March after youth-led protests over corruption and a social-media ban convulsed the country.

New governments usually expect to be judged first on appointments, budgets and promises. Catastrophe replaces that sequence with logistics: which road opens, which family receives a name, which agency can reconcile incompatible lists. The credibility accumulated in an election can disappear at a washed-out bridge.

Nepal contributed roughly 0.1 percent of global greenhouse-gas emissions, according to a World Bank estimate cited by the Associated Press, while remaining acutely exposed to climate hazards. That imbalance is familiar enough to have become diplomatic boilerplate. On the ground it looks less abstract: a government with limited resources must search a terrain altered by a crisis it did almost nothing to cause.

The mourning day supplied a national ritual. The harder work is administrative and repetitive—maintaining missing-person records, housing displaced families, restoring roads before isolation becomes hunger and explaining why some valleys receive help before others. Flags can return to the top of their poles on schedule. Recovery has no such mechanism.

Marvell’s agreement with Google has the dimensions of an investment and the mechanics of procurement.

The chipmaker issued Google a warrant covering as many as 58,970,907 shares at an exercise price of $206.58 each—about $12.2 billion if every share is purchased at that price. But the headline valuation obscures the useful detail. According to Marvell’s regulatory filing, most of the warrant vests through 240 equal tranches, each unlocked by $500 million of qualifying custom-product revenue through fiscal 2033.

In plain English: Google earns more of the right to buy Marvell as Google buys more from Marvell.

The arrangement helps Marvell attach itself to one of the world’s largest consumers of custom silicon. It helps Google cultivate an alternative supplier in an AI infrastructure market where concentration gives incumbent vendors formidable leverage. If Marvell’s value rises, Google can participate in the appreciation partly generated by its own orders.

This is not identical to Google investing $12.2 billion today. A warrant is a right, not the cash purchase implied by casual descriptions of a stake. Exercise depends on vesting, the share price and Google’s choices. Until those conditions align, the spectacular number is a ceiling attached to a long commercial relationship.

The larger development is the fusion of customer and capital provider. Hyperscalers already determine which chip architectures achieve scale. Supplier warrants allow them to capture some of the equity value created by that demand. A vendor wins the contract, but the customer keeps a claim on the victory.

Unitree’s Shanghai debut converted a familiar technology demonstration into a much larger financial claim. Shares of the Chinese humanoid-robot maker rose nearly sixfold in their first session, placing the company at the center of a robotics contest increasingly described as strategic competition between China and the United States.

The public already knows the sales pitch: robots box, run, dance and recover from shoves in clips engineered to travel. Public markets must answer a less cinematic question—whether those abilities become repeatable work at a price below the human or specialized machine they replace.

China has advantages that do not require science fiction. It possesses dense manufacturing networks, a huge domestic market and experience driving down hardware costs. Unitree’s listing gives investors a direct vehicle for betting that these advantages will matter more than the present awkwardness of general-purpose humanoids.

But a debut-day price is not a production statistic. It measures the scarcity of shares and the abundance of expectation. Humanoids still have to cross the difficult distance between controlled demonstrations and long, unsupervised shifts in warehouses, factories, hospitals or homes.

That makes Unitree worth watching for a reason more concrete than its valuation. The company will now publish results into a market that demands cadence: units, revenue, margins, failures and guidance. A dancing robot can repeat a rehearsed sequence. A listed robot company must repeat a business.

Lonnie Bunch helped shepherd the National Museum of African American History and Culture from an argument into a building. He leaves the Smithsonian after a tenure increasingly defined by pressure from the White House and disputes over the institution’s independence.

Museum leadership is often described in the language of acquisition, scholarship and fundraising. Bunch’s later years exposed the other job: maintaining the boundary between public accountability and political command.

That boundary is not clean. The Smithsonian is a public trust with federal support, national reach and enormous authority over which objects receive institutional attention. Elected officials may scrutinize it. But a museum ordered to produce an approved emotional conclusion is no longer curating history; it is manufacturing reassurance.

Bunch’s departure matters because institutional independence is rarely lost in one cinematic act. It is worn down through reviews, funding threats, personnel pressure and demands framed as balance. The administrator becomes the shock absorber, protecting curators from each impact until leaving becomes the final available assertion of control.

His most durable work remains on the National Mall. The museum’s existence changed the physical grammar of American commemoration: Black history was no longer a specialist annex to the national story, but one of its central structures. Buildings can outlast administrations. Whether their interpretive freedom does depends on the people willing to absorb pressure—and on whether the public notices when they are gone.

Netflix plans to report theatrical grosses for six upcoming films, including Greta Gerwig’s *Narnia*. For a company that helped train audiences to regard box-office reporting as an artifact of the old studio system, six is a revealingly small revolution.

The number matters because theatrical performance supplies something streaming platforms struggle to manufacture: a common public event with a visible score. A film can be “most watched” according to an internally defined platform metric, but a ticket sale is legible across companies and decades. It gives agents, filmmakers, rivals and audiences a fact they can compare.

Selective reporting also gives Netflix control. It can use theatrical transparency for projects designed to benefit from spectacle, awards attention or filmmaker prestige without surrendering the opacity surrounding the rest of its catalog.

That makes the experiment less a conversion than a hybrid. Netflix is acknowledging that cinema’s old measurement system still carries cultural authority while preserving the information advantage of streaming everywhere else.

The interesting question is not whether six grosses look impressive. It is whether success makes disclosure contagious. Once audiences and talent receive real numbers for favored films, the absence of numbers for other releases stops looking like innovation and starts looking like editorial judgment.

The modern corporate settlement is praised for efficiency because efficiency is measured from the viewpoint of the docket.

A case ends. Lawyers stop billing. The defendant avoids an admission. The plaintiff receives something tangible. The judge clears space. Every participant can describe the bargain as rational, and usually each is correct.

The missing party is the public record.

When a case concerns a system used by millions—social media, pharmaceuticals, financial infrastructure, industrial pollution—the public needs more than compensation. It needs the instructions: who knew, which warning was ignored, how incentives were arranged and what internal language transformed foreseeable harm into an acceptable metric.

That is why New Mexico’s decision to take its Facebook case toward trial matters beyond the verdict. Trial is an expensive, imperfect knowledge machine. It forces documents into sequences and claims into conflict. It can still fail. But it attempts something a confidential settlement is designed to prevent: a durable account.

We should stop treating every settlement as proof that accountability worked. Sometimes it proves only that uncertainty was successfully priced. A check can make a plaintiff whole; it cannot tell the next regulator where the switch was hidden.

The law will always need settlement. Courts could not function without it. But cases involving society-scale systems deserve a presumption of disclosure, even when the parties compromise. Publish the documents. Preserve the technical record. State which claims remain disputed. Let closure end the lawsuit without erasing the lesson.

A civilization that settles every scandal privately is condemned to rediscover the mechanism publicly.

The corporate euphemism for giving a powerful customer warrants is “alignment.” The less flattering word is tribute.

Marvell’s Google agreement may become an excellent bargain for both companies. Google commits demand to a custom-chip supplier; Marvell gains scale and credibility; the warrant grows valuable only if commercial conditions and the share price cooperate.

But the structure reveals where power sits. The buyer receives the product, the bargaining leverage created by an additional supplier and a claim on the supplier’s equity upside. Ordinary shareholders receive the possibility that all this purchasing will outweigh dilution.

That is not a scandal. It is a map.

AI infrastructure has produced giant buyers whose orders can reorganize a vendor’s future. When the difference between winning and losing one account is measured in factories, years and billions of dollars, procurement starts to resemble private industrial policy. The customer does not merely select the market’s winner. It negotiates a share of the winner before placing the orders that may create it.

Investors should therefore read supplier warrants backward. Do not begin with the enormous notional value. Begin with the purchases required for vesting, the dilution if the shares are exercised and the alternatives the supplier possessed when it signed. “Strategic” often means the counterparty had choices and you did not.

The age of hyperscale computing will generate extraordinary suppliers. It may also ensure that a portion of their extraordinary value belongs, by contract, to the customers they could least afford to refuse.