AI News

OpenAI is doubling down on hardware control, pairing Nvidia GPUs for model training with Broadcom-designed custom chips for inference, according to The Wall Street Journal. The goal? Cut costs, improve…

Amazon is reportedly preparing to lay off as much as 15% of its Human Resources division (PXT), according to Fortune. The move follows CEO Andy Jassy’s June warning that widespread…

Researchers from Nvidia and MIT unveiled QeRL (Quantization-enhanced Reinforcement Learning) — a breakthrough that makes training LLMs with RL up to 1.5× faster, while keeping accuracy equal or better than…

Japan has formally requested that OpenAI cease using copyrighted manga and anime styles in its generative tools, including Sora, after a wave of unauthorized depictions of characters like Mario and…

A new Stanford paper just dropped a bombshell on how we prompt AI models. It’s called Verbalized Sampling, and it proves that aligned models like ChatGPT or Claude aren’t “stale”…

OpenAI and Broadcom just unveiled a massive strategic collaboration to develop and deploy 10 gigawatts of OpenAI-designed AI accelerators — marking a decisive step toward self-built AI hardware at planetary…

This is absolutely wild. The Dutch government has officially taken control of Nexperia, a major Chinese-owned semiconductor firm, in a move that stunned both industry and investors. They didn’t just…

A Korean research team from KAIST has built a new kind of physics-informed AI that can uncover the hidden properties of materials using only a handful of data points —…

NVIDIA has officially trained a 12B-parameter language model on 10 trillion tokens — entirely in 4-bit precision (NVFP4). This is the first-ever stable large-scale 4-bit pretraining run without accuracy loss…

New benchmarks show Claude Sonnet 4.5 achieving a 50% success rate on multi-step software engineering tasks — with an average time horizon of ~1 hour 53 minutes (CI: 50–235 min).…