AI News: Impressive New Model From Unexpected Company

TL;DR

Thinking Machines Labs — founded by ex-OpenAI interim CEO Mira Murati — debuted impressive real-time interaction AI demos including live translation, posture detection, and simultaneous tool calls.

Key Points

• 1.Thinking Machines Labs is the unexpected company with a genuinely impressive new model. Founded by Mira Murati (former OpenAI CTO and brief interim CEO after Sam Altman's firing), the model demos real-time translation, posture detection, context-aware interruption, and simultaneous web browsing — features that feel novel beyond benchmark improvements.
• 2.The Thinking Machines model isn't publicly available yet. Only demos have been shown; a limited research preview is planned in coming months with a wider release later in 2025, distinguishing it from ChatGPT and Claude which lack features like live timekeeping during conversations.
• 3.OpenAI added Codex mobile access, letting users manage coding projects remotely from their phones. The host demonstrated connecting his phone to his Mac Studio via QR code, then querying his personal wiki vault — calling it a 'game changer' for steering AI coding without sitting at a computer.
• 4.Anthropic's Claude subscription changes starting June 15th are being called a massive nerf by power users. Monthly credits burn at expensive API rates, meaning heavy users calculate they'll exhaust credits in hours, then face full API billing — despite Anthropic framing it as expanded access.
• 5.Anthropic beat OpenAI in business adoption for the first time, per Ramp data. Anthropic reached 34.4% of businesses (up 3.8% in April) while OpenAI fell to 32.3% (down 2.9%), alongside Anthropic launching industry-specific Claude packages for legal, financial, healthcare, and small business sectors.
• 6.Google's Android event previewed Gemini-powered features and introduced the 'Google Book' laptop. Highlights include a context-aware Gemini button in Chrome for Android, AI-cleaned spoken text like Whisper Flow, form auto-fill, and a reimagined AI mouse pointer enabling voice-and-hover actions without typing.
• 7.Google IO next week is expected to reveal Gemini 3.2 Flash and a new Spark agent. Gemini 3.2 Flash reportedly handles ~92% of GPT-5.5 coding/reasoning tasks at 15–20x lower cost; Spark is rumored as a 24/7 personal assistant learning from user behavior across connected apps.
• 8.World Labs released an open-source tool that converts a single image into an interactive 3D environment. It generates meshes, physics, lighting, and audio from one photo; usable via Claude Code by cloning a GitHub repo — the host plans to test and potentially make a dedicated video about it.