AI Research

OpenAI’s Noam Brown reportedly says automating AI research is a top priority for agents, a signal with major implications for builders and labs.

OpenAI says 10,000 AI agents solved a $1 million math problem in 88 hours, raising questions about proof, verification, and AI research.

OpenAI says coding agents are changing AI research, but its early evidence leaves key questions about scale, measurement, and independent validation open.

NVIDIA says ICML 2026 papers show open models and open AI infrastructure are now central to research, with implications for robotics, biology and enterprise AI.

New research demonstrates that reformulating the mathematics underlying AI models could significantly reduce memory and storage demands on AI hardware.

Anthropic introduced Claude Science, an AI workbench giving scientists a unified environment for computational research without switching between tools.

VentureBeat explores how hypernetworks generate task-specific model adapters at inference time, sidestepping the costs of fine-tuning and context leaks of RAG.

A new AI-powered brain-computer interface has enabled a speechless ALS patient to hold a full-time job, marking a significant medical AI milestone.

Anthropic will make Claude Fable 5 safeguards visible after criticism that hidden throttling could undermine AI research.

BBC reports Cambridge researchers trialed a vaccine with a key component designed entirely by artificial intelligence.

Google introduced Gemini for Science experiments for hypothesis generation, computational discovery, literature insights, and science skills.

Adaption introduced AutoScientist, an AI tool aimed at automating parts of the research loop behind model training and alignment.

Anthropic published research on translating Claude's internal representations into human-readable text.

WIRED reports a study suggesting even 10 minutes of AI use may negatively affect thinking and problem-solving performance.

MIT researchers report an 81% acceleration for privacy-preserving AI training on everyday devices.

Anthropic ran a week-long internal marketplace with 69 AI agents; more capable models consistently negotiated better outcomes while weaker-model users never noticed.

Meta is launching a dedicated superintelligence research lab while cutting roughly 8,000 employees to redirect resources toward AI ambitions.

Goldman Sachs researchers explain why current AI systems lack a fundamental 'world model' and how solving this gap could reshape the entire AI industry.

A new benchmark reveals that even top AI models drop roughly 50% in accuracy when analyzing complicated charts, exposing a key limitation in visual reasoning.

A chart by METR, a nonprofit AI organization, has become an industrywide obsession as it tracks the rapid development of large AI systems.