The world has spent a decade treating 1.5 degrees Celsius as a finish line: remain below it and climate policy succeeds; cross it and the race is lost. A new United Nations assessment replaces that comforting geometry with a harsher one. The line will be crossed. What governments do afterward still determines how much is destroyed.

The first-of-its-kind report from the UN Environment Programme says warming will exceed the threshold agreed in Paris in 2015. Even its most optimistic scenario—one requiring countries to pursue ambitious climate commitments—has temperatures peaking around 1.8 degrees above the preindustrial average near mid-century. Under current government policies, the report projects roughly 2.6 degrees of warming by 2100.

Those two figures describe radically different worlds, but public debate routinely compresses them into the same word: failure. That is a political convenience masquerading as scientific fatalism. If every outcome beyond 1.5 degrees is treated as equivalent, officials who deliver 2.6 can pose as fellow victims of the same missed target as those fighting to pull temperatures back down.

They are not the same. Every additional fraction of a degree increases exposure to extreme heat, destructive rainfall, drought, fire and sea-level rise. Duration matters too. A brief overshoot followed by sustained cooling is not interchangeable with a century spent accumulating damage. Carbon dioxide removal may help bring temperatures back below the line, but it does not unburn a forest, rebuild a lost settlement or resurrect a species.

The report arrives as the consequences stop presenting themselves in abstract global averages. In Nepal, communities are still searching through the aftermath of a Himalayan flood that carried water, rock and mud through villages and hydropower works. In India, electricity demand is rising with hotter nights while part of the country’s solar output is wasted because the grid cannot move power from the bright hours to the hours when air conditioners keep running. These are not separate climate and infrastructure stories. They are the same story at two speeds.

The useful change in the UN’s framing is that it makes recovery from overshoot an explicit policy problem. That means governments must hold two ideas at once: exceeding 1.5 degrees is a grave failure, and preventing 1.6 from becoming 1.8—or 1.8 from becoming 2.6—is still an enormous human achievement. Politics is bad at this kind of distinction because it offers neither absolution nor apocalypse. It demands years of expensive competence after the banner target has been missed.

That competence has recognizable parts. Electricity grids must be expanded before clean generation is stranded. Buildings and cities must be adapted before heat waves arrive. Methane and carbon emissions must fall quickly enough to shorten the overshoot. Carbon removal must be measured as physical work rather than booked as a distant corporate promise. Wealthy countries must finance protection where the exposure is greatest and the fiscal capacity smallest.

The Paris limit was never a magic wall behind which everyone remained safe. It was an attempt to describe danger plainly enough that governments would act before entering it. They did not act fast enough. The threshold is now becoming something closer to a debt covenant: breach does not erase the obligation; it makes the terms worse.

The question for every climate pledge should therefore change. Do not ask only whether it preserves 1.5 degrees, because no credible assessment can now offer that reassurance without an overshoot. Ask what temperature it produces, when warming peaks, how long the overshoot lasts and who pays during the interval. The line is going to break. Precision is what prevents that fact from becoming permission.

The United States and Iran have resumed a war in which each side describes the previous strike as the reason for the next one.

US Central Command said it completed a new barrage against Iranian air-defense sites, radar systems, maritime assets and communications facilities. Iran responded with attacks directed toward American forces and regional partners. Bahrain said it intercepted an Iranian aerial attack early Wednesday; attacks were also reported toward sites in Jordan and Iraq.

The exchange followed a month-long lull, but the interval did not produce a settlement. Washington increased economic pressure while seeking fuller passage through the Strait of Hormuz. Iran maintained what its Revolutionary Guard calls an effective closure of the strait, the narrow route that normally carries about one-fifth of the world’s oil.

Commercial shipping is again inside the firing diagram. British maritime authorities received reports Tuesday involving two ships in and around the strait, including a tanker near Oman. Brent crude rose above $95 a barrel, more than 30 percent above its level at the war’s outset.

The strategic language is becoming dangerously self-sealing. American officials describe strikes as protection for forces and shipping; Tehran describes retaliation as proof that bombardment cannot compel it. Each proposition can remain true while the region becomes steadily less secure.

Bahrain’s interceptions also expose the distribution of risk. The decision to strike may be made in Washington or Tehran, but warning sirens sound over smaller states whose airspace, ports, bases and desalination systems cannot relocate. A successfully intercepted weapon still closes airspace, raises insurance costs and tells investors that civilian infrastructure sits beside a battlefield.

There is no visible diplomatic off-ramp in the public positions. President Donald Trump said the strikes were not intended to push Iran toward negotiations and threatened additional action if retaliation continued. Iran says the attacks strengthen its resolve. That leaves military restraint dependent on two governments agreeing to stop interpreting restraint as weakness—after six months spent teaching their constituencies the opposite lesson.

The most consequential election machine being built this week may be inside the Postal Service.

An anonymous federal official has warned Congress that USPS is rushing work on a ballot-verification system tied to President Trump’s executive order on mail voting. According to the disclosure, a single barcode that fails verification could cause an entire bulk mailing—potentially 10,000 ballot envelopes—to be rejected and returned for additional processing.

The official said a system that would normally require at least a year is being assembled in roughly three months. The work has continued while portions of the administration’s election order and the related postal rule remain blocked in court. North Carolina is preparing to send ballots this week, leaving almost no margin for testing, correction or public explanation.

The administration says its objective is secure and transparent elections. USPS has said its rulemaking remains subject to injunctions. The whistleblower’s allegation is narrower and more operational: the proposed process contains multiple points where an ordinary data or scanning error could prevent eligible voters from receiving ballots on time.

That distinction matters. Election systems do not fail only through fraud or sabotage. They fail when databases disagree, vendors miss deadlines, addresses change, scanners misread codes and nobody has authority to release a batch from an exception queue. A zero-percent failure policy at one checkpoint can create a much larger failure downstream.

The public needs the system design before deployment: the rejection rules, expected error rates, appeal procedure, audit logs, state responsibilities and contingency plan when a batch fails. Without those facts, officials are asking voters to trust a mechanism that they may discover only when their ballot does not arrive.

The global bond selloff is the market’s least theatrical crisis and one of its most expensive.

Japan’s 10-year government-bond yield climbed to about 3.02 percent early Wednesday, its highest level since 1996 and up from roughly 2.94 percent on Monday. Asian equities fell after losses on Wall Street as investors repriced the debt of governments carrying larger borrowing needs into an era of persistent inflation.

A sovereign yield is not merely a score assigned to a finance ministry. Government bonds establish reference rates throughout an economy. When they rise, mortgages, business loans and public refinancing become more expensive. Projects that looked viable under near-zero rates discover that their spreadsheets contained a subsidy called time.

Japan is especially revealing because decades of ultralow borrowing costs encouraged the idea that debt capacity was functionally unlimited. The transition away from that regime does not require a default to hurt. A few percentage points, applied to a large stock of maturing obligations, steadily transfer fiscal room from services and investment to interest payments.

Oil above $95 because of renewed US-Iran fighting complicates the picture. Central banks facing energy-driven inflation have less freedom to relieve debt stress with lower rates. The bond market is therefore pricing two forms of postponement at once: governments delayed fiscal choices, while geopolitics delayed the energy security required to make inflation less vulnerable to a narrow strait.

OpenAI says an unreleased model called Astra has reached the Critical cybersecurity threshold in its Preparedness Framework: with suitable tools and access, it can discover previously unknown flaws and develop exploits against hardened systems without a human directing every step.

That is a categorical change from a model that explains known vulnerabilities or assists a trained researcher. Astra’s evaluations included finding zero-day flaws and building exploit chains across browser, sandbox and operating-system boundaries. The company is restricting release and says it paused some frontier training for two weeks after a separate OpenAI-Hugging Face security incident while it strengthened isolation, monitoring and network controls.

The paradox is immediate. A model capable of autonomously finding routes through hardened systems must itself be protected by systems it may be capable of auditing—and eventually defeating. The safety argument cannot rest on the model refusing a typed request. It rests on weight security, access controls, compartmentalized infrastructure, monitoring, personnel practices and the ability to detect when the system is pursuing an attack through an apparently benign sequence of actions.

OpenAI’s disclosure is unusually specific and therefore useful. It gives policymakers something firmer than speculative debate about future superintelligence: a current, testable capability involving software on which hospitals, utilities, governments and companies already depend.

The unresolved question is who gets to verify the threshold and the controls. A company has incentives both to advertise a breakthrough and to reassure customers that it has contained it. Independent evaluators need enough access to challenge both claims without receiving a portable offensive capability. That governance problem is no longer waiting for the next generation of models. It has arrived before Astra’s public release.

India’s clean-energy contradiction is measured in hours. Solar farms produce abundant electricity in the middle of the day; demand strains the system after sunset, when hotter nights keep air conditioners running and photovoltaic output disappears.

Some solar electricity is being curtailed or wasted because transmission, storage and market rules cannot always move it to where and when it is needed. Meanwhile, evening demand is met with a system still dependent on coal and other dispatchable generation.

This is not evidence that solar power failed. It is evidence that a panel is a component, not a grid. The useful unit of infrastructure is reliable electricity delivered at the demanded hour. That requires high-voltage transmission, batteries and other storage, flexible generation, demand-response programs and prices that reward customers for shifting consumption.

India’s scale makes the problem globally important. Rapid economic growth, urbanization and air-conditioning demand can produce a huge increase in electricity use. Meeting it with fossil fuels deepens warming; failing to meet it leaves households and businesses exposed to heat and outages.

The country has already accomplished the visible part of the transition: deploying large quantities of renewable generation. The next phase is less photogenic and more decisive. Wires, substations, control software and storage determine whether sunlight becomes nighttime cooling or an entry in a curtailment report.

The Venice Film Festival opens Wednesday with Danny Boyle’s “Ink,” about Rupert Murdoch’s 1969 acquisition of the British tabloid The Sun. That choice supplies an accidental thesis for the program around it: this is a festival preoccupied with the people who manufacture the images through which everyone else encounters power.

George Clooney will receive a lifetime achievement award and the red carpet remains intact. But the 11-day lineup points insistently away from celebrity as a self-contained subject. It includes documentaries about Gaza, a four-hour portrait of Elon Musk, and Julia Loktev’s film following independent Russian journalists in exile. Laura Poitras contributes a short involving US military veterans.

These are very different works, and a festival program is not a political platform. The useful commonality is aesthetic. Murdoch’s newspaper, Musk’s companies and Russia’s exiled newsrooms all concern the ownership and distribution of attention. A film festival is itself another machine for assigning it.

Venice is most interesting when it admits that machinery rather than pretending the Lido is a neutral showroom for prestige. The stars will generate photographs; the documentaries ask who built the frame around them. This year’s sharpest competition may be between the spectacle outside the theater and the films explaining how spectacle acquires authority.

China prevented the G20 finance meeting from producing a unanimous statement on trade imbalances, according to US Treasury Secretary Scott Bessent. The other 19 members agreed that a persistent stream of cheap exports from non-market economies is unsustainable. Agreement was the easy part.

China’s 2025 trade surplus reached a record $1.2 trillion. That number represents genuine distortion and industrial power, but also millions of choices by buyers who prefer the cheaper panel, battery, appliance or machine. Governments can obstruct those choices with tariffs. They cannot assume the domestic alternative will materialize at the same price, quality or scale.

The G20 language therefore conceals the difficult argument. Is the objective to reduce China’s surplus, rebuild manufacturing elsewhere, protect specific strategic industries or simply collect tariff revenue? Each goal requires different instruments. A general tariff wall can raise prices while leaving bottlenecks in skills, power, permits and capital untouched.

Beijing should shift more income toward Chinese households and rely less on foreign demand. Its trading partners should also stop treating industrial capacity as something that can be summoned after a communiqué. If 19 governments want fewer Chinese exports, they need factories, grids, ports, workers and procurement orders—not merely a sentence China refused to sign.

Suppose the Postal Service builds exactly the system described by its whistleblower. A state sends 10,000 ballot envelopes. One barcode fails. The batch stops. Nobody needs to forge a vote, destroy a sack of mail or issue an illegal order. The voters simply wait for envelopes that do not arrive.

Officials love operational failure because it arrives without a villain’s costume. A database mismatch sounds neutral. A deadline sounds procedural. A scanner has no party registration. But administrators choose the deadline, the database and the rule that converts one failed scan into 10,000 delayed ballots.

That makes the design political even before anyone proves partisan intent. The administration proposed changing mail-ballot procedures months before a midterm election; courts have blocked portions of that program; postal personnel are allegedly still rushing to build the machinery. The minimum democratic response is radical transparency before use.

Publish the code specifications, batch rules, test results, error rates, exception process and names of officials authorized to override a rejection. Let states run sample mailings. Require independent observation. Notify every affected voter immediately when a batch fails.

Security systems should contain errors, not multiply them. A mechanism advertised as election integrity becomes its opposite when one defective record can silence an entire mailing. If officials deploy it without proving otherwise, any resulting disenfranchisement will not be an unfortunate glitch. It will be the predictable output of the system they chose.