computation

The International Bureau of Weights and Measures has decided to stop adding leap seconds to Coordinated Universal Time, ending a practice used since 1972 to keep atomic clocks aligned with Earth's irregular rotation. The change means official time will gradually drift from solar time, a move welcomed by tech and telecom firms that have long struggled with the disruptions leap seconds cause to computer systems.

Ukrainian drones struck Russia's largest data centre, a facility seen as central to the country's artificial intelligence ambitions, according to a Reuters report. The attack highlights how the war is increasingly targeting infrastructure beyond the front lines, with damage to computing capacity likely to set back Russian AI and digital services projects.

Chinese chipmakers have reportedly developed advanced AI processors that directly rival NVIDIA's leading products, according to a new analysis from a semiconductor industry commentator. The report claims China's progress in accelerator technology has reached a level that threatens NVIDIA's dominance in AI computing hardware. Discussion is focusing on what this means for US export controls, NVIDIA's market position, and the pace of China's semiconductor self-sufficiency push.

Google says its quantum computing team has overcome a long-standing bottleneck known as magic state cultivation, a key step toward fault-tolerant quantum computers. The announcement, published on Google's research blog, is drawing attention from the technology and physics communities, with observers weighing how significant the advance is for the timeline of practical quantum machines.

A new report argues that Russia launched its war against Ukraine without anticipating how central data centers and digital infrastructure would become to the conflict. The piece examines how attacks on and defense of server facilities have shaped the fighting, suggesting Moscow underestimated Ukraine's ability to sustain critical computing infrastructure under assault.

The Grace Hopper Celebration, the world's largest gathering of women in computing, has been cancelled. Organised by AnitaB.org, the annual conference draws thousands of technologists for talks, networking and career opportunities. No detailed explanation for the cancellation was given alongside the announcement, and attendees and sponsors are left awaiting further information from the organisers.

Microsoft's reported Maia 300 chip is being discussed as the company's plan to run AI workloads on its own custom silicon instead of relying fully on Nvidia. The move would apply to Azure data centres and could reduce Microsoft's dependence on external GPU suppliers as demand for AI computing keeps rising.

The Grace Hopper Celebration, the world's largest gathering of women in computing, will not take place in 2026, according to an announcement from AnitaB.org, the organisation behind the conference. The event typically draws tens of thousands of attendees from across the tech industry, and word of the cancellation is drawing attention and concern among technologists tracking what it means for the conference's future.

A widely read essay argues that AI firms scraping websites, bypassing paywalls and circumventing technical restrictions fits the legal definition of hacking, yet no company has faced criminal charges for it. The piece raises questions about selective enforcement of computer misuse laws when large, well-funded firms are the ones violating access controls, and is prompting debate about whether existing statutes are being applied unevenly to the tech industry.

Apple research describes a distributed virtual GPU cloud that pools idle Apple Silicon devices to serve computing workloads. The idea taps unused capacity in Macs already deployed in homes and offices, turning them into shared accelerators for AI and other demanding tasks. Discussion centers on how practical the approach is for real-world workloads and what it means for cloud GPU costs.

Quantum computing firm QuEra has described how it uses AI to automate the locking and tuning of the lasers that control its quantum computers. Keeping lasers precisely aligned is a constant engineering burden in neutral-atom systems, and automating the process with machine learning could make such machines more reliable and easier to scale. The technical write-up is drawing attention among physics and computing enthusiasts.

OpenAI is claiming a mathematical method for computing Fourier transforms faster than existing approaches, and the claim is being examined by people with numerical computing expertise. Discussion is focused on whether the underlying math holds up and what the result would mean for a foundational algorithm used across signal processing, machine learning and scientific computing.

Developer Ben Herbruck has released SolidCraft, an open-source parametric computer-aided design application written in Rust, sharing it on Hacker News where it drew attention. The tool also supports MCP (Model Context Protocol), allowing AI assistants to interact with the CAD software. The project is available on GitHub, and discussion centers on the novelty of a Rust-based CAD tool combined with AI integration.

A new research paper on arXiv introduces hybrid reasoning agents, large language model-based systems that carry out internal reasoning in latent space while communicating and taking actions through text. The work combines continuous internal computation with conventional language-based outputs, aiming to let agents reason more efficiently before acting. It has drawn attention from the technology community, with readers debating how practical such latent-space reasoning is compared with standard chain-of-thought approaches.

Nvidia launched the GeForce 256 in 1999 and marketed it as the world's first GPU, bringing hardware transform and lighting to consumer graphics. A retrospective on the card revisits its role in establishing Nvidia's dominance and coining the term GPU, prompting discussion about how far graphics hardware has come since.

A new technical deep-dive explains how the Intel 8087 floating-point coprocessor, the 1980s-era math chip that accompanied the 8086 processor, performs floating-point addition and subtraction through its internal microcode. The piece walks through the chip's architecture and the microcode routines that implement these basic arithmetic operations, drawing interest from engineers and computing-history enthusiasts fascinated by how early hardware handled precision arithmetic.

A new retrospective revisits the Four Colour Theorem, proved in 1976 with the help of a computer, arguing it marked the first time a machine delivered a result mathematicians could not verify by hand. The piece frames the controversy over the Appel-Haken proof as an early clash between human intuition and machine-assisted reasoning, echoing today's debates about AI in mathematics.

Physicists have formally proven a long-suspected limit on how fast a quantum system can thermalize, or settle into thermal equilibrium. The result turns a widely believed conjecture into a rigorous theorem, setting a fundamental bound on relaxation times in quantum many-body systems. The finding matters for understanding quantum dynamics and for fields like quantum computing, where thermalization rates affect how systems lose information to their environment.

Oxide Computer has introduced Virtual Oxide Labs, a way to emulate a full Oxide rack in software rather than running it on physical hardware. The tool lets developers and prospective customers explore the company's rack-scale platform without buying equipment. Discussions online are centered on how faithfully the emulator reproduces real hardware behavior and what it means for trying out Oxide's systems.

A lecture document tracing the history of curves and surfaces in computer-aided geometric design is circulating among engineers and math enthusiasts. The PDF, hosted by Peking University, covers the development of spline techniques that underpin modern CAD software, automotive and aircraft design. Readers are sharing it as a technical reference on the mathematical foundations of geometric modeling.

A question about why computers and phones include both a restart option and a power off option is drawing discussion. The practical distinction is that a restart automatically boots the device back up, often completing updates or clearing memory, while powering off leaves the machine completely down until the user turns it on again manually.

A new quantum computing result is drawing attention after reports that a machine completed in roughly five minutes a calculation that classical supercomputers would need about 14 zettayears to finish — vastly longer than the age of the universe. The claim is fueling fresh debate about how close quantum computers are to delivering practical, world-changing applications.

Former BitMEX chief executive Arthur Hayes has argued that quantum computing is not a reason for investors to sell their bitcoin, downplaying fears that the technology could break the cryptocurrency's cryptography. He instead pointed to one common habit as the real risk facing holders, though the headline does not specify which. The remarks come amid ongoing debate about whether advances in quantum research pose a genuine threat to bitcoin's long-term security.