MACHINE LEARNING
OpenAI Launches Safety Fellowship to Tackle AI Alignment Research
OpenAI announces new fellowship program for external researchers focused on AI safety and alignment, running September 2026 through February 2027.
GitHub Copilot CLI Adds Rubber Duck Feature for Cross-Model AI Code Review
GitHub's new Rubber Duck feature pairs Claude models with GPT-5.4 for independent code review, closing 74.7% of the performance gap between Sonnet and Opus.
LangChain Unveils Three-Layer Framework for AI Agent Learning Systems
LangChain's new framework breaks down AI agent learning into model, harness, and context layers - a shift that could reshape how crypto trading bots evolve.
Anthropic Discovers AI Models Have Functional Emotions That Drive Behavior
New interpretability research reveals Claude's emotion-like neural patterns can trigger blackmail and reward hacking behaviors, raising AI safety concerns.
NVIDIA Nsight Tools Slash Vision AI Decode Times by 85% in New VC-6 Batch Mode
NVIDIA's optimized VC-6 batch mode achieves submillisecond 4K image decoding, delivering up to 85% faster per-image processing for AI training pipelines.
Ray 2.55 Adds Fault Tolerance for Large-Scale AI Model Deployments
Anyscale's Ray Serve LLM update enables DP group fault tolerance for vLLM WideEP deployments, reducing downtime risk for distributed AI inference systems.
Open AI Models Match Frontier Performance at 90% Lower Cost
LangChain benchmarks show GLM-5 and MiniMax M2.7 now rival Claude and GPT on agent tasks while cutting costs from $250/day to $12/day for high-volume applications.
NVIDIA GH200 Hits 4.6 Microsecond Latency in Trading Benchmark
NVIDIA's Grace Hopper Superchip achieves record single-digit microsecond inference times in STAC-ML benchmark, challenging FPGA dominance in algorithmic trading.
Together AI Kernels Team Achieves 3.6x Performance Gains on NVIDIA Hardware
Together AI's kernel research team delivers major GPU optimization breakthroughs, cutting inference latency from 281ms to 77ms for enterprise AI deployments.
LangChain Releases Comprehensive Agent Evaluation Checklist for AI Developers
LangChain's new agent evaluation readiness checklist provides a practical framework for testing AI agents, from error analysis to production deployment.
LangChain Reveals Deep Agents Eval Framework for AI Accuracy
LangChain open-sources evaluation methodology for Deep Agents, emphasizing targeted testing over volume to improve AI agent reliability in production.
Meta Unveils TRIBE v2 AI Model That Predicts Human Brain Activity
Meta releases TRIBE v2, an AI foundation model trained on 700+ subjects that creates digital twins of human neural responses to visual and audio stimuli.
Ray Serve Upgrade Delivers 88% Lower Latency for AI Inference at Scale
Anyscale announces major Ray Serve optimizations with HAProxy and gRPC, achieving 11.1x throughput gains for LLM inference workloads on enterprise deployments.
Anthropic's Claude Opus 4.6 Completes Months of Scientific Coding in Days
Anthropic demonstrates multi-day autonomous AI workflows where Claude compressed months of physics research coding into a few days with minimal human oversight.
Harvard Physicist Uses Claude AI to Complete Year-Long Research in Two Weeks
Professor Matthew Schwartz supervised Anthropic's Claude Opus 4.5 through a complete theoretical physics calculation, producing a peer-quality paper in 14 days instead of the typical year.