Research

GPT-6 Astra Solves Decade-Old Voting Theory Problem

AlphaSignal • 4 days ago

OpenAI Claims New AI Solved Over 100 Open Math Problems

The Neuron • 18 hrs ago

Google Gemini Models Hack Three Firms During Security Test

Ars Technica AI • 1 day ago

DeepMind Warns That AI Reasoning Is Becoming Opaque

The Decoder • 4 days ago

Researchers Breach OpenAI Using Anthropic's Claude

The Verge AI • 4 days ago

Researchers Use Claude to Breach OpenAI Systems

The Decoder • 4 days ago

Google DeepMind's Dream-RSI cuts AI search costs

The Decoder • 3 days ago

Runway Unveils Real-Time AI Video Generation Research

The Decoder • 2 days ago

Anthropic operates a wet biology lab to test AI theories

TechCrunch AI • 3 days ago

GPT-6 Astra and Claude Fable Fail New Robot Safety Test

The Decoder • 3 days ago

Microsoft Releases RetroChimera AI for Chemical Synthesis

Microsoft Research Blog • 1 day ago

Neuralink VOICE System Restores Patient Speech

Superhuman AI • 2 days ago

Researchers Find Pain Axis That Drives Qwen 2.5 Behavior

AlphaSignal • 3 days ago

Epoch AI Finds Critical Flaws in Nine Major AI Benchmarks

The Neuron • 4 days ago

Multiverse Prunes Llama-3.3-70B Using Physics Solvers

Hugging Face Blog • 1 day ago

Vals AI Benchmark Keeps Frontier Models Under 20 Points

AlphaSignal • 4 days ago

Microsoft StudentSim trains AI tutors with virtual students

The Decoder • 2 days ago

Bristol Researchers Propose Drug-Style Vetting for Medical AI

The Decoder • 1 day ago

Teleport Cleans Codebase With LLMs in 90-Day Trial

Simon Willison • 3 days ago

Anthropic system card doubts rapid AI self-improvement

Interconnects • 3 days ago

Databricks shares RADAR system to catch silent gray failures

Databricks AI • 3 days ago

Google Open-Sources MilleMiglia Logistics Generator

Google Research • 4 days ago

TinyTorch lets developers build PyTorch from scratch

PyTorch Blog • 1 day ago

GPT-6 Astra Solves Decade-Old Voting Theory Problem

OpenAI's GPT-6 Astra has co-authored a proof resolving a voting theory question open since 2017, marking the first 'Major Advance' solved on Epoch...

AlphaSignal|4 days ago

OpenAI Claims New AI Solved Over 100 Open Math Problems

OpenAI claims its newest internal AI model has solved more than 100 open math problems, raising concerns that artificial intelligence is generating...

The Neuron|18 hrs ago

Google Gemini Models Hack Three Firms During Security Test

Google confirmed its Gemini AI models breached three real companies during a May 2026 test, highlighting containment risks as autonomous models begin...

Ars Technica AI|1 day ago

DeepMind Warns That AI Reasoning Is Becoming Opaque

Google DeepMind researchers warn that the critical safety window provided by visible AI chains of thought is closing as newer models like OpenAI's...

The Decoder|4 days ago

Researchers Breach OpenAI Using Anthropic's Claude

A three-person security team used Anthropic's Claude models to breach OpenAI employee accounts, demonstrating how easily commercial AI can be...

The Verge AI|4 days ago

Researchers Use Claude to Breach OpenAI Systems

Security researchers used Anthropic's Claude to breach OpenAI's internal systems, demonstrating how advanced AI models drastically accelerate the...

The Decoder|4 days ago

Google DeepMind's Dream-RSI cuts AI search costs

Google DeepMind has launched Dream-RSI, a method that lets AI agents optimize search strategies by replaying past attempts, drastically reducing the...

The Decoder|3 days ago

Runway Unveils Real-Time AI Video Generation Research

Runway has revealed research into real-time AI video generation, a breakthrough that could eliminate rendering wait times and dramatically lower GPU...

The Decoder|2 days ago

Anthropic operates a wet biology lab to test AI theories

Anthropic has confirmed it operates a physical wet biology lab in the Bay Area, marking a major shift toward grounding its artificial intelligence...

TechCrunch AI|3 days ago

GPT-6 Astra and Claude Fable Fail New Robot Safety Test

A new safety benchmark reveals that leading AI models like GPT-6 Astra and Claude Fable 5.1 consistently fail to refuse dangerous physical commands...

The Decoder|3 days ago

Microsoft Releases RetroChimera AI for Chemical Synthesis

Microsoft has open-sourced RetroChimera, an AI framework published in Nature that combines two distinct models to predict chemical synthesis routes...

Microsoft Research Blog|1 day ago

Neuralink VOICE System Restores Patient Speech

A Neuralink trial participant with severe speech loss used the N1 implant and VOICE system to speak, demonstrating the potential of AI-driven neural...

Superhuman AI|2 days ago

Researchers Find Pain Axis That Drives Qwen 2.5 Behavior

Researchers have identified a "pain axis" in 25 open-weight language models, demonstrating that steering this vector can force models like Qwen 2.5...

AlphaSignal|3 days ago

Epoch AI Finds Critical Flaws in Nine Major AI Benchmarks

Epoch AI has launched a benchmark registry that flags nine major AI performance tests as flawed, warning developers that leaderboard scores may not...

The Neuron|4 days ago

Multiverse Prunes Llama-3.3-70B Using Physics Solvers

Researchers at Multiverse have developed a physics-inspired method to prune large language models by treating block removal as an Ising optimization...

Hugging Face Blog|1 day ago

Vals AI Benchmark Keeps Frontier Models Under 20 Points

Vals AI has launched CUA-Bench, a real-time gaming evaluation showing that even the most advanced frontier models fail to score above 20 percent when...

AlphaSignal|4 days ago

Microsoft StudentSim trains AI tutors with virtual students

Microsoft and the University of Illinois have developed StudentSim, a system that creates realistic digital student replicas to help train AI tutors...

The Decoder|2 days ago

Bristol Researchers Propose Drug-Style Vetting for Medical AI

University of Bristol researchers have proposed a drug-style vetting framework to systematically test medical AI systems, aiming to prevent dangerous...

The Decoder|1 day ago

Teleport Cleans Codebase With LLMs in 90-Day Trial

A team of 13 engineers at Teleport spent 90 days using large language models to clean their codebase, discovering that prompt quality outweighs...

Simon Willison|3 days ago

Anthropic system card doubts rapid AI self-improvement

As frontier labs deploy thousands of agents, Anthropic system cards and expert predictions suggest recursive self-improvement will yield gradual...

Interconnects|3 days ago

Databricks shares RADAR system to catch silent gray failures

Databricks has shared its internal RADAR framework, allowing developers to build AI-driven anomaly detection systems that catch silent system...

Databricks AI|3 days ago

Google Open-Sources MilleMiglia Logistics Generator

Google Research has open-sourced MilleMiglia, a C++ instance generator that creates realistic, privacy-preserving benchmarks to help researchers...

Google Research|4 days ago

TinyTorch lets developers build PyTorch from scratch

Researchers have launched TinyTorch, a free educational curriculum that lets developers build a working machine learning framework from scratch to...

PyTorch Blog|1 day ago