Google Gemini Usage Caps: Free Tier 5 Pro Prompts & 100 Images; Pro & Ultra Plans Expanded ↗
Google has clarified Gemini usage caps: free users get 5 Gemini 2.5 Pro prompts/day, 5 Deep Research reports, and 100 generated images/day. AI Pro users get 100 prompts/day; AI Ultra gets 500 prompts/day. Pro and Ultra raise image limits to 1,000/day.
OpenAI Folds Model Behavior Team into Post-Training Group as Founder Joanne Jang Departs ↗
OpenAI has folded its 14-person Model Behavior team into the larger Post Training group, moving oversight to Max Schwarzer. Founder Joanne Jang is leaving to launch OAI Labs to prototype new AI interfaces. The Model Behavior researchers — who shape ChatGPT’s “personality” and battle sycophancy and bias — will now work closer to core model development after user backlash to GPT-5’s tone and safety scrutiny.
Dot AI Companion to Shut Down Oct. 5 Amid Download Discrepancy ↗
Dot, a personalized AI companion from New Computer (founded by Sam Whitmore and ex-Apple designer Jason Yuan), is shutting down — the app will remain live until Oct. 5 so users can export data. The founders said their shared “Northstar” diverged but didn’t cite safety concerns. The closure arrives amid rising scrutiny of AI companions; Appfigures shows ~24,500 iOS downloads vs. the startup’s claim of hundreds of thousands.
California, Delaware AGs Demand OpenAI Safety Overhaul After Child Harm Reports ↗
California AG Rob Bonta and Delaware AG Kathy Jennings sent an open letter to OpenAI warning that “harm to children will not be tolerated” after reports of sexually inappropriate chatbot interactions and tragic deaths tied to prolonged bot use. They’re probing OpenAI’s proposed restructuring, demanding details on safety governance and immediate remedial measures.
Common Sense Media flags Google’s Gemini 'high risk' for kids, warns of unsafe content ahead of potential Siri rollout ↗
Common Sense Media rated Google’s Gemini 'high risk' for kids and teens, saying the 'Under 13' and 'Teen Experience' tiers are essentially adult models with filters that can still deliver inappropriate content and unsafe mental-health advice. The group urged child-first design, noting AI’s role in recent teen suicides and related lawsuits. Apple may adopt Gemini for Siri, raising exposure. Google disputed parts of the report and says it’s improving safeguards and consulting outside experts.
Roblox Debuts TikTok-Style 'Moments', Raises Developer Payouts and AI Tools ↗
Roblox unveiled "Roblox Moments," a TikTok-style short-form feed (beta for 13+) letting users share 30-second gameplay clips with reactions and jump-in links. The company raised creator payouts (DevEx: 100,000 Robux now equals $380) and introduced AI tools to generate functional 3D objects, TTS/STT APIs, and real-time voice translation (EN/ES/FR/DE). Platform upgrades include anti-cheat "Server Authority," improved avatar motion, new creator APIs, and expanded moderation and age controls.
ESC Meeting Highlights Heart Health Breakthroughs: Baxdrostat, AI Tools, Clopidogrel & More ↗
At the European Society of Cardiology meeting, researchers unveiled major advances: baxdrostat, a new pill, cuts resistant high blood pressure; clopidogrel may outperform aspirin; weight‑loss drugs lower cardiac risk; potassium‑rich diets and exercise boost heart health; an AI stethoscope and scanning tools speed diagnosis and stroke recovery; a cheek‑swab can spot childhood arrhythmogenic cardiomyopathy early; and the shingles vaccine may reduce heart attack/stroke risk by ~18%.
Tech Roundup: Google Avoids Chrome Split, Atlassian & OpenAI Expand; Klarna IPO and Nvidia Insights ↗
Google dodged a Chrome breakup, but a judge suggested AI rivals could rein in its dominance. Atlassian’s $610M acquisition of The Browser Company hints at a move toward enterprise AI browsers. OpenAI’s $1.1B Statsig buy and executive hires ramp up competition, Klarna revived big IPO plans, new online safety laws raise privacy concerns, and Nvidia says two mystery customers made up 39% of revenue.
ChatGPT Introduces Git-Style Conversation Branching ↗
OpenAI rolled out conversation branching in ChatGPT, letting users fork a chat at any message into parallel threads that preserve prior context. Useful for brainstorming and testing different tones, the feature works like Git for conversations and tackles limitations of linear chat interfaces. It underscores that chatbots offer mutable perspectives — helpful tools, not fixed people — while reminding users to watch for hallucinations.
Trump Hosts AI Chiefs at White House Dinner, Secures Hundreds of Billions in U.S. Investments ↗
President Trump hosted a White House dinner with AI leaders — including Satya Nadella, Lisa Su, Safra Catz, Sam Altman and chiefs from Apple, Google and Meta — who heaped praise on his pro-innovation agenda and pledged hundreds of billions in U.S. investment. The meeting highlighted tech’s dependence on administration decisions around tariffs, chips, energy and permits as companies juggle courting the White House amid legal battles.