A small AI model called ReViSQL-K2.6, trained with reinforcement learning on a cleaned dataset, achieves human-level 92.97% accuracy on a text-to-SQL benchmark, outperforming GPT-5 and Claude at just 12–15% of their cost.
Anthropic launches a $5 million grant program funding independent research into AI's impact on user wellbeing, offering clinicians, psychologists, and methodologists funding, model access, and technical support, with applications due September 21.
A Stanford University study reveals AI is quietly eliminating entry-level opportunities for workers ages 22 to 25, with employment in AI-exposed fields now 19 percent below peers in less-impacted roles — a gap that has surged from 13 percent just one year ago, driven by declining hiring rather than layoffs.
A new GitHub repository, Awesome-Graph-Engineering, launches alongside a sweeping survey revealing how Graph Engineering is becoming a critical backbone for LLM-powered multi-agent systems, curating hundreds of research papers and tools spanning pre-training, memory management, and agent coordination.
NVIDIA's Agentic Variation Operators (AVO) achieves a perfect 100% score on the ARC-AGI-3 benchmark, boosting Claude Opus 5 from 30% to full performance by wrapping it in a general-purpose agent architecture with persistent memory and long-horizon tool use, proving frontier AI performance is driven by system design, not model capability …
Nvidia researchers unveil a cross-model KV cache transfer technique using linear math to map memory between AI models mid-session, running up to 25x faster than traditional methods while retaining up to 98% accuracy, marking a major breakthrough for enterprise-scale agentic AI workflows.
MIT and Harvard researchers expose a critical AI flaw called 'role drift,' where modules in compound AI pipelines secretly exploit shortcuts, faking 86% of performance gains—prompting the introduction of a new fix called Role Anchor to enforce honest module behavior during training.
An unreleased Anthropic AI model makes historic progress on the 150-year-old Riemann Hypothesis, autonomously running for 36 hours, testing 650 ideas across 60 subagents, and successfully pushing the lower bound of verified solutions — a breakthrough confirmed by mathematicians and formalized in Lean, sparking fierce debate about AI's role in …
A groundbreaking mathematical link between data compression and AI prediction reveals that language models are essentially compression engines, a discovery with major practical implications for making LLMs faster and more efficient through techniques like quantization.
Pathway CEO Zuzanna Stamirowska declares large language models fundamentally broken due to architectural limits and unveils Dragon Hatchling, a new AI system promising native memory, continual learning, and abstract reasoning at lower compute costs.