Javier Alarcón

The Xring O3, Xiaomi’s new flagship processor, scored 3,945 points on Geekbench single-core and 15,221 on multi-core. Its 44 MB of cache and 21 execution ports point to a trend toward massively parallel ARM cores, though the technical community questions the legitimacy of the figures.

A security researcher discovered that Microsoft Paint and the Photos app embed a server-issued GUID as an invisible watermark in every image generated with local artificial intelligence. The feature is independent of the visible Copilot logo and cannot be turned off from the settings.

UC San Diego announced it will test an experimental power technology to supply AI data centers. The initiative responds to the growing electricity demand required by large-scale AI model training and inference.

Bluesky’s iOS app swaps the Follow button for its butterfly logo the instant the system detects a screenshot. The mechanism, found in the app’s open source code, repurposes an iOS feature meant to hide passwords.

The CH32V003, a ten-cent RISC-V microcontroller, is at the center of a response from an embedded engineer in Trinidad and Tobago to the harshest article the architecture has received in years. The debate exposes a real tension between design elegance and hardware access outside the United States and Europe.

A technical critique of RISC-V’s design, published by engineer Dmitry.GR, breaks down why the open ISA’s interrupt handling is more costly than that of a low-end ARM Cortex-M0 core. The analysis counts cycle by cycle how the lack of automatic register stacking forces developers to write assembly routines before reaching the C code.

Oxide Computer built three Kubernetes integrations, Rancher, Omni, and Cluster API, following the real problems its customers ran into when deploying clusters. The process included a Talos Linux bug that prevented reading Oxide’s user-data and an improvised temporary fix before KubeCon.

NVIDIA launched Nemotron 3.5 Lightning, a 30-billion-parameter MoE model optimized for high-volume AI agents. Alongside it comes NeMo Switchyard, an open source library that routes each task to the most suitable model based on cost, latency, and quality.