Isaac ROS 5 brings agents into robotics development
NVIDIA updates its GPU-accelerated packages built on ROS.
26 stories
Research
Science, robotics, and questions still open.
Mistral connects open models with regional infrastructure
The company expands regional endpoints and outlines longer-term compute plans.
Gemini Robotics 2 extends intelligence across the whole body
Google focuses on movement, dexterity, and cooperation between robots.
Cosmos-H-Dreams generates simulations for surgical robotics
NVIDIA describes a faster way to explore action-conditioned surgical scenes.
Grabette records robot data without starting with a robot
Pollen Robotics proposes an open handheld system for manipulation demonstrations.
VoiceEQ looks at what speech benchmarks miss
Hume AI researchers examine the human quality of voice interaction.
Robostral Navigate uses one camera to get around
Mistral introduces a model for following navigation instructions in physical spaces.
LeRobot 0.6 closes more of the learning loop
World models, reward models, and evaluation tools join the robotics toolkit.
Leanstral 1.5 makes formal proofs more accessible
Mistral releases an Apache-licensed model for Lean 4 proof engineering.
Claude Science gathers research tools into a workbench
Anthropic emphasizes inspectable outputs and access to computation.
ScarfBench tests agents on Java migrations
IBM evaluates modernization across a project rather than an isolated function.
LifeSciBench asks AI the kinds of questions researchers face
OpenAI introduces expert-developed and checked life-sciences evaluations.
Strands and LeRobot connect the robotics learning loop
AWS describes an agent workflow spanning data, training, and hardware.
GPT-Rosalind extends its research capabilities
OpenAI updates its specialized life-sciences model.
Mistral explores AI for physical simulation
The company positions models as another tool for engineering design.
Gemini for Science follows more of the research process
Google introduces experimental environments around scientific work.
AlphaEvolve searches for better algorithms
Google describes applications in scientific and infrastructure work.
GPT-Rosalind specializes in life-sciences research
OpenAI introduces a reasoning model for biological research workflows.
Gemini Robotics-ER 1.6 looks more closely at the physical world
The update focuses on spatial reasoning before a robot acts.
Anthropic’s next models need tomorrow’s computing capacity
A Google and Broadcom agreement points toward additional TPU infrastructure from 2027.
EVA evaluates the conversation, not just the transcript
ServiceNow researchers look at both task completion and the human experience of voice AI.
Measuring intelligence takes more than one score
Google DeepMind proposes a cognitive framework for evaluating AI progress.
LeRobot 0.5 connects more of the robotics workflow
Hugging Face expands hardware, control policies, and data tooling.
Gemini Deep Think takes aim at research problems
Google updates its specialized reasoning mode for science and engineering.
Beyond the parameter count
A new architecture paper is a reminder that bigger is only one way a model can change.
How to read an AI benchmark
Start the year with a better question: what, exactly, did the test measure?