MACHINE LEARNING
OpenAI Drops IH-Challenge Dataset to Harden AI Against Prompt Injection Attacks
OpenAI's new IH-Challenge training dataset improves LLM instruction hierarchy by up to 15%, strengthening defenses against prompt injection and jailbreak attempts.
Together AI Upgrades Fine-Tuning Platform With Vision and Reasoning Support
Together AI adds tool calling, reasoning traces, and vision-language fine-tuning to its platform, with 6x throughput gains for 100B+ parameter models.
OpenAI Launches GPT-5.4 Mini and Nano for High-Volume AI Workloads
OpenAI releases GPT-5.4 mini and nano models with 2x faster speeds and dramatically lower costs, targeting coding assistants and agentic AI systems.
NVIDIA Dynamo 1.0 Ships With 7x Inference Boost for AI Data Centers
NVIDIA releases Dynamo 1.0, an open-source inference OS adopted by AWS, Azure, Google Cloud, and major AI companies. Claims 7x performance gains on Blackwell GPUs.
NVIDIA Drops Nemotron 3 Super With 5x Throughput Gains for AI Agents
NVIDIA releases Nemotron 3 Super, a 120B parameter open model delivering 5x higher throughput for agentic AI with a 1M-token context window.
LangChain Defines Agent Harness Architecture for AI Development
LangChain's new framework breaks down how agent harnesses turn raw AI models into production-ready systems through filesystems, sandboxes, and memory management.
DeepMind Marks 10 Years Since AlphaGo Changed AI Forever
Demis Hassabis reflects on AlphaGo's decade-long impact, from Nobel Prize-winning protein folding to AGI development. Here's what it means for AI's future.
NVIDIA AIConfigurator Slashes LLM Deployment Time With 38% Performance Gains
NVIDIA's open-source AIConfigurator tool optimizes LLM serving configurations in seconds, delivering 38% throughput improvements for disaggregated AI inference deployments.
OpenAI Finds AI Reasoning Models Cant Hide Their Thinking - A Win for Safety
OpenAI's new CoT-Control benchmark reveals frontier AI models struggle to obscure their reasoning chains, reinforcing monitoring as a viable safety layer.
Google Unleashes Gemini 3.1 Pro and AI Music Tools in February Blitz
Google's February 2026 AI rollout includes Gemini 3.1 Pro with 2x reasoning gains, Lyria 3 music generation, and Nano Banana 2 image tools for developers.
OpenAI Abandons SWE-bench Verified After Finding 59% of Failed Tests Were Flawed
OpenAI reveals major contamination issues in SWE-bench Verified benchmark, showing frontier AI models memorized solutions and tests rejected correct code.
Anthropic Brings Software Testing Rigor to AI Agent Skills
Claude's skill-creator update adds evals, benchmarks, and A/B testing for non-engineers building AI agent skills. Here's what it means for the ecosystem.
NVIDIA Run:ai Delivers 2x GPU Utilization Gains for AI Inference Workloads
NVIDIA benchmarks show Run:ai platform doubles GPU utilization while cutting latency 61x for enterprise AI deployments running NIM inference microservices.
NVIDIA NVFP4 Training Delivers 1.59x Speed Boost Without Accuracy Loss
NVIDIA's NVFP4 4-bit training format achieves 59% faster AI model training than BF16 while matching accuracy on Llama 3 8B benchmarks, per new research.
Anthropic Study Reveals Users Skip Critical Checks on AI-Generated Code
New research from $380B-valued Anthropic shows users are 5.2% less likely to verify AI outputs when creating artifacts, raising questions about automation risks.