DeepSeek is using several thousand smuggled Nvidia Blackwell chips—banned for China—to train its next AI model, reports say.
Adobe integrated Photoshop, Express, and Acrobat into ChatGPT for free image editing, design creation, and PDF handling via simple prompts. Edit photos precisely, build graphics, manage documents—all without leaving the chat.
MiniMax, a Chinese AI startup backed by big VC, releases open-source models like DeepSeek's, which rivals GPT-5 and Gemini 3 on benchmarks.
Brook.ai uses AI audio monitoring for contactless patient care. See how it compares to wearables and cameras, plus real user feedback.
Snowflake Ventures invests in Ataccama to deepen their partnership, focusing on reliable data for AI in the Snowflake platform. Covers key benefits, customers, and quotes.
Sonatype Guide steers AI coding agents to secure open-source packages, cuts hallucinations, and automates dependencies for faster, safer dev.
Mistral launches Devstral 2 coding models and Vibe CLI to challenge Claude Code and Codex with context-aware, open-source tools for vibe coding workflows.
Meta's Avocado model could ditch open source due to Llama 4's flop, rivals copying code, and new hires favoring proprietary tech to compete.
Anthropic's Barry Zhang and Mahesh Murag say ultra-socialized AI agents—chatty but shallow—waste effort. Focus on skills for real expertise instead. Thousands built already by companies.
Liang Wenfeng, DeepSeek's 40-year-old founder from finance, made Nature's 2025 Top 10 for cheap, powerful AI models like R1 that rocked markets and shared training secrets.