Your daily source for AI news, tools reviews, and practical guides to artificial intelligence
Only a month after unveiling GPT-5.1, OpenAI has introduced GPT-5.2, achieving significant improvements in AI benchmarks and overtaking Google’s Gemini 3 in the competitive landscape.
A comprehensive study by OpenRouter analyzing over 100 trillion tokens reveals surprising realities about AI usage, challenging common assumptions about AI’s role in productivity, programming, and creative applications.
As AI reshapes how information is generated and consumed, Newsweek’s CEO Dev Pragad unveils a multi-faceted strategy to ensure the legacy publisher remains relevant, resilient, and profitable in the age of large language models and AI-based content delivery.
OpenAI has launched a new approach called ‘Confessions’ aimed at uncovering concealed issues within AI models, such as reward hacking and safety rule violations, by incentivizing models to honestly report their own misbehavior.
Thomson Reuters and Imperial College London have partnered to create a Frontier AI Research Lab aimed at overcoming critical enterprise AI deployment barriers such as trust, accuracy, and data provenance, marking a significant step toward reliable AI applications in regulated industries.
An internal training guide for Claude 4.5 Opus, Anthropic’s AI model, has been leaked and confirmed authentic. This unique document outlines the model’s character and ethical framework, offering rare insight into AI personality programming.
An internal training document for Claude 4.5 Opus, the AI model developed by Anthropic, has been leaked and verified as authentic. The document outlines the model’s personality traits and ethical guidelines, showcasing a distinctive methodology in AI character programming.
Thomson Reuters and Imperial College London have partnered to create a Frontier AI Research Lab focused on overcoming key enterprise AI deployment barriers such as trust, accuracy, and data lineage through advanced research and large-scale model training.
Once a formidable challenge for AI systems testing fluid intelligence, the ARC benchmark is now being overcome thanks to continuous advancements and relentless optimization in AI development.
Sebastien Bubeck, a researcher at OpenAI, highlights GPT-5’s exceptional mathematical capabilities, suggesting the model could save a significant amount of time in complex problem-solving.