Anthropic: Claude Science, an AI workbench for scientists (Jun. 30, 2026)
Claude Science is an AI workbench that brings common research tools, data connectors, and preconfigured scientific skills into one reproducible workspace, with auditable code, figures, and a reviewer agent.
OpenAI: OpenAI and Broadcom unveil LLM-optimized inference chip (Jun. 25, 2026)
OpenAI and Broadcom unveiled Jalapeño, a custom inference accelerator for LLMs that promises much better performance per watt, lower latency, and full-stack optimization. Built in nine months with OpenAI models, it will scale to gigawatt deployments with partners.
OpenAI: Daybreak: Tools for securing every organization in the world (Jun. 25, 2026)
OpenAI is expanding Daybreak to democratize machine-speed patching, launching an updated Codex Security plugin, GPT-5.5‑Cyber, a partner program, and Patch the Planet to help open-source projects.
Anthropic: Project Fetch: Phase two (Jun. 18, 2026)
Project Fetch tested Claude with a robodog; Opus 4.7, alone, was about twenty times faster than humans on shared tasks. It still struggled with precise beach-ball retrieval, showed limited low-level control, and needs more research, testing, and evaluation.
Anthropic: Making Claude a chemist (Jun. 5, 2026)
Opus 4.7 matched or outperformed classical tools on shifts, excelled at peak splitting and consistency, and showed promise for structure elucidation.
WSJ: Amazon to Invest Additional $13 Billion in India by 2030 (Jun. 25, 2026)
Amazon will invest an additional $13 billion in India by 2030 to expand AWS data centers, provide AI chips, managed AI services, and add fulfillment centers, raising its India commitment to over $88 billion.
NY Times: OpenAI and Broadcom Unveil Custom A.I. Chip Design (Jun. 24, 2026)
OpenAI and Broadcom unveiled a custom AI chip, Jalapeño, to run models like ChatGPT and cut dependence on Nvidia and AMD. OpenAI plans to use these chips at scale.
NY Times Opinion: There’s One Clear Reason Americans Are Gloomy About A.I. (Jun. 24, 2026)
Americans are unusually pessimistic about A.I., unlike most countries, because U.S. labor-market institutions—weak unemployment benefits, employer-linked health care, uneven protections—make job loss catastrophic.
Simon Willison: The AI Compass (Jun. 30, 2026)
The AI Compass is a political-compass quiz with 29 questions about AI, AI ethics, and personal stances, assigning one of 30 archetypes like “The Garage Tinkerer”.