The PyTorch and Nebius teams joined forces to accelerate the pre-training of DeepSeek-V3 on modern GPUs, and the results exceeded expectations.
Why the new competitive barrier in the world of AI isn't algorithms or data, but the ability to skillfully build agent management systems.
Higress has been accepted into the CNCF as a Sandbox project – an open AI gateway that manages traffic, ensures security, and monitors models.
AI: Events
When 31% of the Cache Just Vanishes: The Story of a Silent Bug Deep Within GPU Code
Technical context • Development
AI21 engineers spent weeks hunting down mystical glitches during model training, only to find the culprit in two characters of code at the GPU level.
AI: Events
How a Japanese Company Teaches AI to 'Feel' Metal: ARUM's Approach to Precision Manufacturing
Products
Japanese startup ARUM is converting decades of knowledge from expert craftspeople into data, enabling AI to replicate their precision on an industrial scale.
Inception Labs has introduced Mercury 2 – a diffusion language model that operates quickly and affordably, paving the way for a new approach to creating AI agents.
Researchers have developed a new simulation method that enables more accurate measurement of signals from dark matter particles in semiconductor detectors at the single-quantum level.
The Allen Institute has introduced MolmoWeb, an open-source web agent. It navigates browsers visually, much like a human, and outperforms many proprietary competitors.
JetBrains has unveiled a platform for managing AI agents in development. The system is designed to transform the chaotic use of neural networks into a single, transparent, and manageable process.