AI Updates on 2025-06-28

Gemini 2.5 Pro is now available again in the free tier of Google's API, allowing developers to build applications without cost @OfficialLoganK
Google Gemini app introduces scheduled actions for Pro and Ultra users, allowing automated recurring tasks like daily calendar summaries and weekly event searches @GeminiApp

Meta continues aggressive AI talent acquisition, hiring four more researchers from OpenAI as part of their superintelligence lab expansion @AndrewCurran_
Developer reports spending $4,600 annually on AI engineering tools including Claude, ChatGPT, Cursor, and Devin, compared to $100k on human engineers, arguing it's essential for efficiency @clairevo
Pedigree from prestigious companies and warm referrals are becoming the primary pathways into hiring funnels, indicating a tighter job market in tech @GergelyOrosz
Developer demonstrates building a complete UI library with 20 components in 48 hours using Claude Code, suggesting significant productivity gains for those adopting AI coding tools @deedydas

Ethan Mollick emphasizes expanding AI ethics beyond model creators to include user responsibility, stating that "cheating is still cheating" and "misusing a tool is still misusing a tool" regardless of AI capabilities @emollick
Gergely Orosz criticizes Anthropic's messaging about AI replacing workers, calling them "the least responsible lab" for promoting mass unemployment while claiming to be responsible AI developers @GergelyOrosz
Amanda Askell from Anthropic reflects on the importance of ensuring AI training data represents the kind of AI systems we want to create, noting concerns about current collective efforts @AmandaAskell
Simon Willison warns about a phishing attempt targeting AI users, highlighting security vulnerabilities in the AI ecosystem @simonw

Ethan Mollick demonstrates using o3 for personalized shopping advice by simply taking a photo and asking "is this a good deal?", suggesting AI could transform retail experiences @emollick
Perplexity launches WhatsApp integration that can convert daily tasks like news summaries into audio podcasts @AravSrinivas
Neuralink demonstrates brain-computer interface allowing users to control cursors, play games like Mario Kart and Call of Duty, and operate robotic arms for writing using only thoughts @deedydas
Nevada desert facility uses 805 retired EV batteries to power a 2,000 GPU AI data center, creating North America's largest microgrid and demonstrating sustainable AI infrastructure @TechCrunch

Hugging Face releases the Seamless Interaction Dataset, the world's largest in-person conversation dataset with over 4,000 hours of two-person interactions and 4,000+ unique participants @jffwng
Stanford research reveals extensive evidence of close relationships between computer vision research and surveillance technology, published in Nature @stanfordnlp
Simon Willison advocates for "context engineering" as a more accurate term than "prompt engineering," emphasizing that previous model responses are key to the process, not just user prompts @simonw
Hamel Husain identifies common AI-generated writing patterns including overuse of phrases like "The key insight" and "Remember... the goal is not to X but Y," suggesting need for fine-tuning to improve writing quality @HamelHusain