NerdNews

Intel's DeepMath is a lightweight agent that generates small Python scripts to support LLM problem-solving, reducing errors and output length. It's built on Qwen3-Thinking and fine-tuned with GRPO, and is available on GitHub and Hugging Face.

AMD unveiled its new Ryzen AI 400 Series processor at CES 2026, featuring 12 CPU Cores and 24 threads, allowing for 1.3x faster multitasking and 1.7x faster content creation. The company also announced the AMD Ryzen 7 9850X3D gaming-focused processor and the latest version of its Redstone ray tracing technology.

Uber, Lucid Motors, and Nuro have revealed a production-intent robotaxi at CES 2026. The vehicle is based on the Lucid Gravity SUV and features high-resolution cameras, solid-state lidar sensors, and radars. It will be tested on public roads before a commercial service launches in the San Francisco Bay Area later this year.

Nvidia released a new stack of robot foundation models, simulation tools, and edge hardware at CES 2026, signaling its ambition to become the default platform for generalist robotics. The company introduced new open foundation models, including Cosmos Transfer 2.5, Cosmos Predict 2.5, and Cosmos Reason 2, as well as Isaac GR00T N1.6, a next-gen vision language action model. Nvidia also launched Isaac Lab-Arena, an open-source simulation framework, and deepened its partnership with Hugging Face to make robotics development more accessible.

Boston Dynamics has partnered with Google's AI research lab, DeepMind, to develop its next-generation humanoid robot, Atlas. The partnership aims to integrate DeepMind's AI foundation models with Boston Dynamics' Atlas robots, enabling them to interact with humans more naturally and perform tasks like assembly and parts sequencing. Atlas has 56 degrees of freedom, human-scale hands with tactile sensing, and can lift up to 110 pounds. The robot is expected to be deployed in Hyundai's factory by 2028.

This article discusses the importance of handling secrets securely in Terraform, including using environment variables, storing secrets in secure external storage, and implementing Role-Based Access Control (RBAC). It also covers integrating secrets management tools like HashiCorp Vault and AWS Secrets Manager with Terraform, and using automated secrets detection tools like GitGuardian's ggshield.

Brendan McLoughlin shares a detective’s framework for solving cloud infrastructure failures, discussing the 'usual suspects' in the network stack and strategies for building robust runbooks and using tools like Honeycomb and OpenTelemetry

The article discusses the importance of treating latency as a first-class product concern and provides architectural lessons for building sub-100ms APIs. It highlights the need for a latency budget, async fan-out, caching hierarchies, and circuit breakers to achieve low latency. The article also emphasizes the importance of observability, instrumentation, and distributed tracing in maintaining low latency.

AWS has announced two new capabilities for S3 Tables: intelligent tiering, which automatically optimizes costs based on access patterns, and replication support, which maintains consistent Apache Iceberg table replicas across AWS Regions and accounts. Intelligent tiering automatically moves data to the most cost-effective tier, while replication support creates read-only replicas that stay in sync within minutes.

Google published a guide outlining eight essential design patterns for multi-agent systems, including sequential pipelines, coordinator/dispatcher, parallel fan-out/gather, and human-in-the-loop architecture. These patterns help developers design scalable and reliable agentic applications using the Agent Development Kit.

LG's CLOiD robot can perform various household chores, including loading the washer and dryer. It was showcased at CES 2026, where it demonstrated its ability to load a towel into a washing machine in about 30 seconds. The robot has animated eyes, a masculine-sounding voice, and can extend its arm to interact with objects. It can also bring items to people, such as a bottle of water. LG claims CLOiD can perform tasks like grabbing milk from the fridge and toasting a croissant in the oven.

Nvidia's CEO Jensen Huang announced that the company's next-generation AI superchip platform, Vera Rubin, is in full production. The new chips will cut the cost of running AI models to about one-tenth of Nvidia's current leading chip system and can train certain large models using roughly one-fourth as many chips. The Rubin system includes six different chips and will be used by partners such as Microsoft and CoreWeave.

Google DeepMind's Gemini AI is being integrated into Boston Dynamics' humanoid robots, such as Atlas, to enable them to navigate unfamiliar environments and manipulate objects. The collaboration aims to test Gemini-powered Atlas robots at Hyundai's auto factories, marking a significant step towards using humanoids in manufacturing.

Nvidia launched Alpamayo, a family of open source AI models, simulation tools, and datasets for training physical robots and vehicles. Alpamayo 1, a 10 billion-parameter chain-of-thought model, allows autonomous vehicles to reason through complex driving situations and explain their decisions. The models are available on Hugging Face, and Nvidia is also releasing an open dataset and simulation framework for validating autonomous driving systems.

Kodiak AI collaborates with Bosch to develop a system for giving standard big rigs autonomous driving capabilities, aiming to scale its self-driving tech for the truck masses. Bosch will supply hardware components, including sensors and vehicle actuation components, to enable the next generation of autonomous trucking.

Agentic AI is a new class of identity that behaves like humans but operates with machine scale and persistence, posing significant identity and security risks that CISOs must address through lifecycle management and visibility

Google TV's new update introduces Nano Banana, a text-to-image model, allowing users to generate images and videos on their TV. The feature will be available on certain TCL devices first and will expand to other Google TV devices in the coming months. Users can search their Google Photos library and apply AI effects to their photos, creating cinematic slideshows and original media directly on their TV.

The 2026 BMW iX3 will feature Alexa+, a next-generation voice assistant powered by generative AI. This partnership between BMW and Amazon aims to provide a seamless and natural conversation experience for drivers, allowing them to handle complex requests and actions across different services.

Teradar has unveiled its first terahertz-band vision sensor, called Summit, designed for high-performance in any weather. The sensor is meant to fill the gap left by legacy radar and lidar sensors, and is expected to start shipping in 2028. It has the potential to enable partial or full autonomy features in vehicles.

A study by OpenAI found that 60% of American adults have used AI like ChatGPT for health or healthcare in the past three months, with 40 million people worldwide asking ChatGPT healthcare-related questions daily. OpenAI sees this as an opportunity to shape policy and expand in the medical industry, despite concerns about the potential for medical misinformation.