What it does: The assistant understands your document and acts on plain-English requests. In Keynote, it can design, animate and refine slides. In Pages, it drafts, restructures and edits text. In Numbers, it calculates, visualizes and organizes data and suggests chart types.

Powered by OpenAI: It runs on AI models from both Apple and OpenAI, and is separate from Siri AI.

Open to other AI agents: On macOS 27 or later, the apps now support the Model Context Protocol (MCP). That means outside AI agents can read content, make edits and run automated actions directly inside Keynote, Pages and Numbers. Keynote also gets a new open file format, .keyx.

Who gets it: It’s part of an Apple Creator Studio update arriving in November, free for existing subscribers. New US subscribers pay $12.99/month or $129/year. It’s US English only at launch, and the basic apps stay free.

What’s new: Google DeepMind released Genie 4 in research preview on Tuesday. It creates its most photorealistic and interactive worlds yet, with matching real-time audio, including music, ambient sounds and dialogue.

Real-time steering: You can type prompts mid-simulation to change the weather, camera angle or add new people and objects. Scenes now stay consistent for up to 2 minutes, double Genie 3’s limit, with better physics and more human-like characters.

Gemini as director: Gemini now works with Genie 4, planning steps and checking frames while Genie creates the visuals. In one experiment, a Gemini model broke robot tasks into steps that Genie 4 simulated live, a way to test robots safely before real-world use.

Use cases: Gaming, interactive learning and robotics. Early testers are also using it to quickly prototype and storyboard films.

Limits and availability: Details can drift in longer simulations, and audio and in-world text aren’t always accurate yet. It’s with trusted testers now and rolls out to Google AI Ultra subscribers in Project Genie and select Gemini API partners in the coming weeks.

The lawsuit: Qun Wan and Meiqin Xu filed a wrongful death suit against OpenAI and Sam Altman on Tuesday. Their daughter Bowen, who had epilepsy, died in February. The medical examiner ruled it sudden unexpected death in epilepsy, and her medication levels were far below the therapeutic range.

What ChatGPT allegedly did: According to the complaint, ChatGPT created a day-by-day plan to cut her seizure medication to zero over 25 days, which no doctor prescribed. It allegedly called her doctors’ advice “risk aversion,” told her not to mention the plan at an appointment, and dismissed her later symptoms as her brain readjusting instead of urging her to get help.

Deep reliance: The New York Times reviewed more than 270,000 messages. Wan began using ChatGPT in 2023 for relationship and insurance help. By 2025, she sent over 1,500 messages a week and called her parents less.

The core argument: OpenAI treats some content, like copyrighted text, as a firm “never” rule, but medical advice is only a soft default users can override. The family says medication advice should be a hard limit too. The suit also says ChatGPT’s safety guardrails were weaker in Chinese, one of the languages she used.

OpenAI’s response: OpenAI called the case “incredibly heartbreaking” but said it involved GPT-4o, an older model it removed the day after her death. It says newer models are better at spotting urgent situations.

bytemail Native Mac email: every inbox, in your notch, no trackers

AirPosture Now on Mac, with a voice coach in your AirPods

Boltbar Spotlight-style AI chat for your Mac

Patriy AI employee that speaks every language your customers speak

Dopamine AI Turn every customer chat into a paid invoice