OpenAI's GPT-5 assisted in solving a long-unsolved problem from mathematician Paul Erdős's 1992 list, highlighting AI's role in pushing math forward. A new report details this and other research wins.
Ai2's OLMo 3 family introduces fully open models with visible reasoning and high efficiency, perfect for local runs via Ollama. Includes Base, Think, and Instruct variants outperforming competitors on benchmarks.
Project Prometheus is Jeff Bezos' AI startup focused on engineering and manufacturing tools, co-led with Vik Bajaj and backed by $6.2 billion.
US approves 35,000 GB300 chips each for Saudi HUMAIN and UAE G42, boosting regional AI data centers for training and inference.
Agent Laboratory uses LLM agents to automate literature reviews, experiments, and reports, cutting research costs by 84% and matching state-of-the-art results, per a new EMNLP 2025 paper.
OpenAI's ChatGPT 5.1 brings adaptive thinking and a chattier style, but tests against Gemini 3 reveal wins in logic and stumbles in creativity and coding.
Dynatrace and TDK integrate Amazon Bedrock AgentCore for better AI observability and edge monitoring, opening doors for dependable agents in biomedical research and healthcare.
AppsFlyer launches eight products to unify data, boost AI automation, and enable privacy-safe collaboration, evolving into a Modern Marketing Cloud.
Microsoft's new platform tames AI agents by treating them like employees, with tools for tracking, securing, and scaling bots across your company.
Google's Antigravity integrates with Gemini 3 to let AI agents handle coding tasks autonomously, from editing to browser control, with verifiable outputs and memory for smarter workflows.