The US Space Force has terminated OCX, a ground control system for military GPS satellites, after RTX Corporation (formerly Raytheon) failed to deliver working software despite a 16-year, $8 billion effort. The program, awarded in 2010 for completion by 2016, was meant to handle new signals from GPS III satellites and enable jam-resistant M-code for military use. Last year's delivery still failed integrated testing. Col. Stephen Hobbs, commander of the unit operating the GPS constellation, stated that "problems across a broad range of capability areas would put current GPS military and civilian capabilities at risk." The Space Force will instead upgrade the decades-old legacy control system, which has already been partially modernized to support some GPS III features. A $105 million contract to Lockheed Martin for ground upgrades supporting next-generation GPS IIIF satellites, awarded earlier this month, signaled the shift before the official cancellation.
OCX stands for Operational Control System, the next-generation ground software for commanding GPS satellites. GPS III and GPS IIIF are newer satellite generations with improved accuracy and anti-jamming capabilities. M-code is an encrypted military signal designed to resist spoofing and jamming.
Victory Giant, a Guangdong-based manufacturer of printed circuit boards for AI and high-performance computing, jumped 57% in its Hong Kong trading debut on Tuesday. The company raised HK$20 billion (US$2.6 billion) in one of Asia's largest IPOs this year, with retail investors oversubscribing 431 times. Victory Giant supplies Nvidia and was promoted to tier-one supplier status in 2024 after entering its H-series AI accelerator supply chain the prior year. The company now leads global PCB revenue in AI/high-performance computing with 13.8% market share, up from seventh place and 1.7% in 2024. The strong debut reflects investor appetite for hardware infrastructure plays amid the AI build-out, even as questions persist about US-China technology restrictions and supply chain realignments.
Victory Giant (胜宏科技) specializes in high-density interconnect PCBs for data center and AI workloads. The company established a dedicated division for this segment in 2019, positioning it to capture demand from Nvidia's AI accelerator ramp. Hong Kong has seen renewed IPO activity after several quiet years.
California Attorney General Rob Bonta has unsealed a 16-page document detailing how Amazon allegedly orchestrated price-fixing with vendors including Levi's, Scotts, and Hanes. The evidence, filed with the Supreme Court in February as part of a 2022 lawsuit, shows Amazon directing brands to raise prices at competing retailers like Walmart, Target, and Home Depot — or remove discounted products entirely — to ensure Amazon maintained the lowest visible price. In one example, Amazon asked Levi's to push Walmart to raise Easy Khaki prices from below $30 to $29.99; Levi's confirmed compliance. Another shows Amazon directing Scotts to have competitors raise prices for three days before Prime Day. Bonta called the written evidence "explicit and egregious." The state seeks a preliminary injunction to halt the practices while litigation proceeds. Amazon did not immediately respond to requests for comment.
California's 2022 lawsuit against Amazon alleges anticompetitive practices in its marketplace operations. The newly unredacted filing reveals direct communications between Amazon and vendors coordinating price increases at rival retailers. The case tests whether Amazon's most-favored-nation pricing policies violate antitrust law.
A detailed technical analysis by cryptographer Filippo Valsorda clarifies that Grover's algorithm — the quantum search method often cited as halving symmetric key security — does not in practice threaten AES-128 or SHA-256. The common misconception that quantum computers force a move to 256-bit keys for 128-bit security stems from misinterpreting how Grover's algorithm parallelizes. Unlike classical brute force, which is "embarrassingly parallel," Grover's algorithm loses its quadratic speedup when distributed across multiple quantum computers. Partitioning the search space of size N across 2^16 machines only reduces per-machine work by 2^8, not 2^16. A decade-long attack on AES-128 would require sequential oracle invocations at microsecond gate speeds — physically implausible at scale. NIST and other standards bodies maintain that existing symmetric cryptography needs no changes for post-quantum security. The finding matters because it redirects limited migration resources toward actual vulnerabilities: public-key systems broken by Shor's algorithm.
Grover's algorithm provides quadratic speedup for unstructured search, often mischaracterized as simply halving key lengths. Shor's algorithm breaks RSA and elliptic-curve cryptography, driving the actual post-quantum migration urgency. AES-128 and SHA-256 remain secure against known quantum attacks.
Researchers at UC Berkeley's BAIR lab have developed GRASP, a gradient-based planner that makes long-horizon control practical in learned world models. The method addresses three core failures in existing approaches: deep computation graphs from backpropagating through long rollouts, brittle gradients through high-dimensional vision encoders, and poor exploration in deterministic optimization. GRASP lifts trajectories into virtual states for parallel optimization across time, injects stochasticity directly into state iterates for exploration, and reshapes gradients to preserve clean action signals while avoiding unstable "state-input" paths through vision models. The technique enables planning over extended horizons where standard gradient descent collapses into local minima or explodes. World models — differentiable simulators learned from data — are increasingly capable of predicting visual sequences across tasks, but using them for control has remained fragile. GRASP bridges this gap, potentially accelerating applications in robotics and autonomous systems where long-horizon reasoning matters.
World models are learned dynamics models that predict future states given actions, enabling planning by optimization. Long-horizon planning fails with standard methods due to gradient instability and local optima. GRASP modifies the optimization landscape to make gradient-based planning viable at scale.
Forbes embedded its ForbesPredict platform — a token-based prediction market launched in January to boost engagement amid declining search traffic — into coverage of a Louisiana mass shooting where eight children were killed. The article invited readers to wager tokens on whether Congress would pass gun safety legislation before year-end, with a hint available for 10 tokens and predictions costing 100. The same interface offered odds on whether Trump would pardon Ghislaine Maxwell. ForbesPredict uses no real money; users earn tokens for status and "gameplay advantages." Chief Innovation Officer Nina Gould described the goal as deepening engagement rather than chasing scale. The integration of predictive gaming with breaking news about child deaths prompted immediate criticism. The incident illustrates the friction between publisher survival strategies and editorial boundaries as AI-driven search changes reshape traffic economics.
ForbesPredict is a gamified prediction platform built with Axiom, designed to increase time-on-site as search referrals decline. Unlike Polymarket or Kalshi, it uses no real money. The platform launched in January 2026 as part of Forbes' response to AI platforms answering queries that previously drove traffic to publishers.
Posit has released ggsql, an alpha implementation of the grammar of graphics that lets users describe visualizations directly in SQL syntax. The tool integrates with Quarto, Jupyter notebooks, Positron, and VS Code, accepting standard SQL queries that funnel results into visualization specifications marked by a VISUALIZE clause. Users map data columns to aesthetic properties and layer geometric elements — points, lines, smoothers — with the compositional flexibility that made ggplot2 influential in R. The approach eliminates context-switching between data manipulation and visualization languages. Because the visualization query builds on the SQL result, incremental refinement is natural: add a color mapping, swap a geom, adjust scales without rewriting data pipelines. The implementation uses DuckDB as its backend in demonstrations, though the SQL portion passes through to whatever database the user specifies.
The grammar of graphics, formalized by Leland Wilkinson and popularized by ggplot2 in R, separates visualization into data, aesthetic mappings, geometric objects, and coordinate systems. ggsql adapts this framework to SQL workflows, targeting analysts who want visualization expressivity without leaving their query environment.
Universal basic income is resurfacing in policy conversations as tech companies and political figures promote it as a response to AI-driven labor displacement. The renewed interest comes with a specific framing: rather than redistributing existing wealth, proposals now center on "public wealth funds" that would channel returns from AI infrastructure and natural resources directly to citizens. OpenAI CEO Sam Altman has advocated this approach, and similar ideas appear in Elon Musk's rhetoric about abundance. The political logic is transparent — a check in the mail might defuse backlash against automation and the companies building it. The piece examines why this iteration differs from earlier UBI advocacy, which focused on poverty reduction and administrative efficiency rather than technological transition. It also notes the skepticism warranted when the same firms driving labor disruption propose the remedies, and the tension between UBI's individualistic cash transfer and traditional progressive priorities around services and collective provision.
UBI proposals have circulated since the 1960s, with pilot programs in Finland, Kenya, and Stockton, California producing mixed results. The "public wealth fund" variant ties payments to resource or technology returns rather than taxation, drawing on models like Norway's sovereign wealth fund or Alaska's Permanent Fund Dividend.
Evolutionary psychology explains why lay economic beliefs frequently contradict formal theory, argues this piece. Humans evolved in small, tight-knit groups with zero-sum competition between coalitions — not in anonymous markets with comparative advantage and gains from trade. The mismatch produces predictable errors: treating international trade as theft rather than mutual benefit, viewing prices as moral judgments rather than scarcity signals, and resisting market mechanisms that expand cooperation beyond kin and clan. Anthropologist Pascal Boyer and political scientist Michael Bang Petersen's "folk-economic beliefs" framework suggests these intuitions are adaptive responses to ancestral problems, not ignorance or irrationality. The article traces specific policy errors — minimum wage assumptions, rent control support, tariff enthusiasm — to their cognitive origins in coalition psychology and zero-sum thinking. The implication is that economic education faces deeper barriers than information gaps: it must work against evolved mental structures that made sense for most of human history but mislead in modern market contexts.
The "folk-economic beliefs" research program applies evolutionary psychology to explain systematic deviations from neoclassical economic predictions. It distinguishes between genuine irrationality and adaptive heuristics misapplied to novel contexts. The framework suggests economic policy debates often reflect cognitive defaults rather than evidence evaluation.
The 130th running of the Boston Marathon concluded Monday with the traditional finish line on Boylston Street. The 26.2-mile course wound through eight Massachusetts cities and towns, from Hopkinton to downtown Boston. Reporter Alex Ashlock, covering the race for over two decades, filed from the finish area. The event maintains its status as the world's oldest annual marathon and one of six World Marathon Majors, with qualification standards that remain among the most demanding for non-elite runners. This year's race occurred under clear conditions with temperatures favorable for performance. The Boston Marathon holds particular significance in the region's civic calendar, drawing spectators across the route and generating substantial economic activity. Security protocols established after the 2013 bombing remained in effect, with visible law enforcement presence along the course.
The Boston Marathon began in 1897 and has run annually since, with 2020's cancellation the only interruption. The World Marathon Majors series includes Boston, London, Berlin, Chicago, New York City, and Tokyo. Qualifying times vary by age and gender, with the 2026 standard at 3:00:00 for men 18-34 and 3:30:00 for women 18-34.
