AI news on papers, experiments, academic findings, technical deep-dives. Showing stories before September 5, 2026.
Debating LLM Self-Referentiality And Logic
EvalDetectBench Measures LLM Evaluation Awareness
OpenAI Launches GPT-6 Astra
Google Maps Complete Fruit Fly Brain
Google Releases Gemini 3.8 Flash
Gemini 3.8 Flash Adds Thinking Levels
Debating LLM Self-Referentiality and Logic
New Benchmark Detects AI Evaluation Awareness
Apple Unveils REFACTOR-VLA For Robot Motor Programs
HyperWorld Optimizes Textual World Model Serialization
Anthropic Publishes Claude System Prompts
CMRE Symposium To Explore AI In Diabetes
Google Releases August 2026 AI Updates
BenchMIRT Exposes LLM Benchmark Data Contamination
Claude Writes 180,000 Lines of Direct2D Code
HyperWorld Optimizes State Serialization For World Models
Domain-Adaptive Pretraining Boosts Propaganda Detection
ARC-AGI Performance Hits 44% For 67 Cents
BenchMIRT Exposes LLM Benchmark Data Contamination
HyperWorld Optimizes Textual World Model Serialization
Claude Writes 180,000 Lines of Direct2D Code
AI Detects Heart Disease From ECGs Instantly
ARC-AGI Score Hits 44% For 67 Cents
Arctic Expeditions Warn Against AI Overfitting
BenchMIRT Exposes LLM Benchmark Data Contamination
Claude Fable 5.1 Doubles Science Benchmark Score
HyperWorld Optimizes Textual World Model Serialization
AI Detects Heart Disease From ECGs Instantly
ARC-AGI Score Hits 44% For 67 Cents
Claude Fable 5.1 Doubles Science Benchmark Score
BenchMIRT Exposes LLM Benchmark Data Contamination
ARC-AGI Score Hits 44% For 67 Cents
iMML Package Solves Incomplete Multi-Modal Data
AI Learning Patterns Challenge Intelligence Definitions
Australian Teens Use AI For News
OpenAI Splits ChatGPT Work Into Cloud And Local
AI Writing Tools Impact Workplace Voice
ChatGPT Work Splits Into Cloud And Local
OpenAI Splits ChatGPT Work Into Two Versions
Google AI Automates Planetary Prediction Models
Apple Finds LLMs Fail Bayesian Belief Updates
Mathematicians Debate AI Superhumanity At OpenAI
ChatGPT Work Splits Into Cloud And Local
Apple Finds LLMs Fail Bayesian Belief Updates
Mathematicians Debate AI Superhumanity At OpenAI
OpenAI Splits ChatGPT Work Into Two Versions