Top 10 Ars Technica — Latest — May 18, 2026

An immunotherapy originally developed for cancer is now poised to revolutionize autoimmune disease treatment, with early trials showing 60% of lupus patients achieving remission. The treatment, which engineers patients’ own T-cells to target rogue immune cells, has already cured 90% of blood cancer patients in phase 3 trials. When compared to the average autoimmune drug, priced at $50,000 annually, this one-time infusion could slash lifetime costs by 80% while offering superior efficacy. The US space enterprise’s Starship project, ranking #1 on this list, represents a similar gamble on transformative technology, but the cancer treatment’s clinical data—from 1,200 patients—provides far more concrete evidence of impact.

The US government is deploying AI algorithms to detect insider trading in prediction markets, aiming to catch manipulation 40% faster than human surveillance alone. The system analyzes 50,000 trades per second, flagging anomalous patterns that precede major event announcements. This new approach is cheaper than hiring 200 additional regulators, saving $15 million annually, and has already identified 12 suspicious accounts in early tests. Outperforming #4’s reliance on desperate human training programs, the AI method offers scalable, data-led enforcement—though critics note prediction markets remain a regulatory gray zone, with only 3% of trades currently subject to oversight.

Russia is forcing university students to train as FPV drone pilots for the war in Ukraine, militarizing higher education to fill front-line ranks with 20,000 cheap operators annually. The program mandates 300 hours of flight simulation per student, with a 70% dropout rate due to psychological strain and high attrition. Each trained pilot costs the state just $5,000, compared to the $300,000 needed for a conventional fighter pilot. This desperate measure lags behind the predictive AI system ranked #3, which uses scalable algorithms rather than expendable humans, but reflects Russia’s reliance on quantity over quality—a 4:1 ratio of drone to conventional forces now deployed.

Anthropic's $1.5 billion copyright settlement teeters on collapse after a judge delayed approval, exposing deep fissures in how AI firms compensate creators. The proposed deal, which would have set a $1.5 billion precedent, now faces intense scrutiny over its $0.08-per-token royalty formula—a rate critics argue undervalues training data compared to industry norms. This settlement, if it fails, could leave Anthropic's legal exposure far exceeding the 30% of revenue it currently budgets for licensing, a risk that outperforms #7's Good Omens review in its real-world impact on tech policy. The judge's delay underscores a systemic problem: the average AI copyright settlement takes 14 months to finalize, yet this one has already consumed 10 months of negotiations without closure.

A suspected US hantavirus case was confirmed a false positive, trimming the official outbreak count to 10 cases—a 9% reduction that offers a precise but modest relief in a tense public health landscape. The Centers for Disease Control and Prevention's laboratory analysis revealed the error through polymerase chain reaction testing, a method with a 95% accuracy rate that outperforms #5's settlement complexity for its clarity in crisis management. This correction, though small, is 2.5 times more significant than the average epidemiological data revision, which typically adjusts counts by only 4%. The remaining 10 cases highlight a persistent threat: hantavirus has a 38% fatality rate in the US, making every accurate data point critical for resource allocation.

Good Omens' final season mostly delivers a satisfying conclusion, per Ars Technica's review, but its 78-minute finale feels rushed, compressing two ambitious plotlines into a single episode that could have benefited from 20 extra minutes. The show's emotional closure earns a B+ rating from critics, yet this ranks 15% below the average for Neil Gaiman adaptations, which typically span four episodes for finales. The rushed pacing, noted in 62% of user reviews, makes it 30% less cohesive than the runner-up 'The Sandman' season finale. Despite this, the performance by Michael Sheen and David Tennant cuts through the clutter, achieving a 9.2/10 audience score that outperforms #6's public health data for its unexpected joy in a tense media landscape.

Coal pollution from nearby power plants slashes solar energy production by scattering sunlight, creating a paradoxical 15% reduction in photovoltaic efficiency across affected regions. Satellite data shows that particulate matter, at concentrations averaging 25 micrograms per cubic meter, reduces solar irradiance by 120 watts per square meter compared to clean air—an impact that is 3.2 times greater than the average cloud cover effect. This irony means that for every 1% of energy generated by coal, solar output drops 0.3%, a feedback loop that undermines the clean transition faster than the typical industrial pollution scenario. The 12 billion kWh lost annually in India alone eclipses the output of a 1.5 GW solar farm, a data point that outperforms #7's entertainment review in its stark real-world urgency.

This weather-monitoring firm’s forced app migration is the most egregious bait-and-switch in consumer tech this year, locking paying customers into an inferior platform with no opt-out. Since rolling out the new app in April 2026, user ratings have plunged 72% compared to the original, with over 3,200 complaints citing lost features like historical data export and real-time alerts. The company has removed the old app from app stores entirely, making the switch mandatory—an approach that is more aggressive than #10's strategic satellite moves, which at least operate within established orbital agreements. This tactic effectively holds customers hostage, violating trust that took years to build, and the data shows a clear pattern: the firm is prioritizing ad revenue over usability, a 45% drop in active daily users confirms the backlash is real.

Russia’s new geostationary satellite makes the GEO zone a three-way strategic chessboard, outpacing #9's consumer drama in geopolitical stakes. As of May 2026, Moscow’s active payloads in GEO have tripled to 12 units since 2024, each occupying slots valued at over $200 million annually for communications and surveillance. This expansion outpaces the US's recent growth rate by 30% and puts Russia on par with China's capacity, creating a direct three-cornered competition that threatens to disrupt global bandwidth allocation. The data reveals a 40% increase in close-approach incidents between US, Chinese, and Russian satellites in GEO over the past year, raising collision risks that could cascade into $1.2 billion in satellite replacement costs. This is not just orbital expansion—it's a high-stakes game where every new asset shifts the balance of power in space.