LAUNCHES NVIDIA ProRL Agent Decouples RL Training from Multi-Turn LLM Rollouts 8/10 4 min read 2 months ago
RESEARCH Google Research Publishes TurboQuant Two-Stage LLM Compression System 8/10 4 min read 2 months ago
RESEARCH Karpathy’s AutoResearch Agent Runs 700 Experiments in Two Days, Cuts GPT-2 Training Time 11% 7/10 2 min read 2 months ago
RESEARCH GPT-5.2 and Claude Opus 4.6 Both Go Silent on ‘Ontologically Null’ Prompts, Study Finds 8/10 2 min read 2 months ago
RESEARCH Study Finds AI Users Adopt Incorrect Answers 73% of the Time When Models Err 8/10 2 min read 2 months ago
RESEARCH MoonshotAI’s AttnRes Replaces Fixed Residuals with Depth-Aware Attention 7/10 4 min read 2 months ago