Official release from AI companies

OpenAI has introduced PaperBench, a benchmark designed to assess AI agents' capabilities in replicating leading AI research from ICML 2024, including understanding, coding, and executing experiments.

OpenAI has launched PaperBench, a benchmark designed to assess AI agents' ability to replicate leading AI research from ICML 2024. This includes understanding papers, coding, and conducting experiments, evaluated through detailed rubrics.

Anthropic announces Claude for Education, a new initiative partnering with universities to integrate AI into higher education, featuring a dedicated learning mode for students.

Anthropic has launched Claude for Education, partnering with universities to integrate AI into higher education. This new learning mode is available for students at select institutions and Pro users with .edu emails.

Google DeepMind announces the publication of DreamerV3 in Nature, an AI that learns to solve control tasks by predicting future outcomes. Notably, it autonomously found diamonds in Minecraft without human data.

Microsoft reflects on its technological journey, highlighting 15 key milestones from the inception of floppy disks to the launch of Copilot+ PCs, showcasing its evolution and future direction.

Microsoft reflects on its technological journey, highlighting 15 key milestones from the inception of floppy disks to the launch of Copilot+ PCs, showcasing its evolution and future direction.

OpenAI has launched the new 'Monday' voice feature for ChatGPT, accessible to all users. To use it, update the app and select 'Monday' from the voice picker in voice mode.

OpenAI has introduced a new voice feature in ChatGPT, enhancing user interaction with a more natural conversational experience.

OpenAI has introduced a new voice feature in ChatGPT, allowing users to access it by opening voice mode and selecting the voice picker in the top right corner.

Google DeepMind announces the rollout of Gemini 2.5 Pro, an experimental AI model, to all Gemini users. The team emphasizes their commitment to making this advanced model widely accessible as they continue to enhance its capabilities.

Google DeepMind highlights the launch of Gemini 2.5 Pro, showcasing its advanced reasoning, coding capabilities, and multimodal strengths, including a 1M token long context, in a detailed breakdown by product leads.