Artificial intelligence and machine learning breakthroughs, models, and research.
26 articles
The race to build the most capable AI system just became a cybersecurity crisis waiting to happen.
A new initiative aims to tackle one of the most overlooked vulnerabilities in modern AI systems by hardening the open-source dependencies that power them.
Google's latest open-weight model family arrives with benchmarks that challenge proprietary systems head-on, raising serious questions about the future of closed AI development.
A new MIT-led technique called CompreSSM shrinks state-space AI models during training itself, cutting compute costs without sacrificing accuracy.
Muse Spark handles complex reasoning and multimodal tasks but abandons the open weights approach of previous generations.
A massive npm packaging error revealed 500,000 lines of proprietary source code powering Anthropic's flagship AI assistant.
OpenAI put its UK Stargate data center on hold, citing high electricity prices and regulatory uncertainty while it reassesses where to place long term AI compute.
ByteDance's most powerful AI video model has started reaching international creators after months of legal battles, geo-restrictions, and a delayed global rollout.
Just one month after the 30-second Lyria 3 debut, Google's upgraded Pro model pushes AI music generation to three full minutes with structural awareness across six platforms simultaneously.
ByteDance's top-ranked AI video model just went fully public via fal.ai, giving developers unrestricted access to cinematic image-to-video generation.
Cursor cloud agents don't just write code and hope it works — they film themselves doing it, then attach the proof to the pull request.
Google's Gemini just replaced static AI answers with live, manipulable simulations you can spin, adjust, and explore inside the chat window.
