Cursor Research has open-sourced Mixture-of-Kittens (MoK), a specialized training megakernel designed for Mixture-of-Experts (MoE) models. This megakernel integrates all communication and computation steps within an MoE layer into a single, deterministic process.
Earlier reports mentioning Nvidia
NVIDIA has introduced a sophisticated pipeline for auditing the security of AI skills, leveraging its open-source tool, SkillSpector. This comprehensive workflow integrates with LangGraph, a framework for building stateful, multi-actor applications, along with YARA rules for pattern matching, SARIF…
Thinking Machines Lab has released Inkling-Small, an open-weights Mixture-of-Experts (MoE) model featuring 276 billion total parameters and 12 billion active parameters. This new model is approximately a quarter of the size of its predecessor, Inkling, which boasts 975 billion total…
NVIDIA's NeMo team has released Molt, a new PyTorch-native framework for agentic reinforcement learning, designed to streamline the research process by reducing the complexity of algorithm modification. The framework aims to be compact enough for researchers to understand its entirety…
NVIDIA's Transformer Engine is designed to accelerate the training of transformer-based artificial intelligence models by leveraging a combination of fused GPU kernels, BF16 (Brain Floating Point 16) computation, and hardware-aware FP8 (8-bit floating-point) execution.
Samsung's semiconductor division is experiencing an exodus of engineers who are migrating to its South Korean rival, SK Hynix. This talent drain is primarily driven by substantial financial incentives offered by SK Hynix, which is capitalizing on record profits generated…
Engineers at Samsung's semiconductor division are increasingly seeking employment with rival SK Hynix, driven by substantial financial incentives. One engineer, identified only as Lee, described a demoralizing situation at Samsung where employees are actively applying to SK Hynix, even sharing…