Google DeepMind Introduces Genie 4
Google DeepMind has released Genie 4, a world model that generates realistic, interactive environments along with synchronized audio.
- Genie 4 generates more photorealistic environments than its predecessor, Genie 3, and is natively multimodal, producing real-time audio that matches the simulated world, including music, ambient noises, and interactive dialogue.
- Users can edit scenes in real time through text prompts, changing the weather or camera angle or adding people and objects without restarting the simulation.
- The model can maintain a consistent scene for up to two minutes, double Genie 3’s limit, and offers improved physics and more humanlike characters.
- Gemini is more closely integrated with Genie 4, helping plan actions and evaluate frames while Genie generates the environment. In a robotics research experiment, a Gemini model broke tasks into steps that Genie 4 simulated in real time, suggesting a way to test robotic scenarios before deploying them in the physical world.
- Early use cases include interactive games, educational experiences, robotics simulations, and film prototyping.
Genie 4 still has limitations. Objects and scene details can drift during longer simulations, and Google is working to improve audio accuracy and the legibility of text inside generated environments. It is currently available as a research preview to “trusted testers,” and Google plans to expand access in the coming weeks to Google AI Ultra subscribers as well as to select Gemini API and Google AI Studio partners.
ChatGPT Can Now Access Your iCloud Mail and Calendar
OpenAI has added iCloud Mail and Calendar integrations to ChatGPT, allowing the assistant to search emails, manage messages, and schedule or update events.
- The Calendar integration can read, schedule, and update events that appear in Apple’s Calendar app or other apps using iCloud Calendar.
- The Mail integration can search and read emails, manage messages and drafts, send emails, and look up iCloud contacts.
- iCloud Mail uses an Apple-provided email address that ends in @icloud.com, @me.com, or @mac.com. ChatGPT can only access your iCloud Mail, not other accounts in the iOS Mail app, such as Gmail or Outlook.
- Users need to create an app-specific password to connect both iCloud Mail and iCloud Calendar.
The integrations are available across all ChatGPT plans on the web, desktop, and iOS.
Create Downloadable Files with Gemini
Step 1: Open Gemini and clearly describe what you want it to create. You can ask for documents, slideshows, or spreadsheets. If needed, upload relevant files or images, such as handwritten notes, screenshots, scanned documents, etc.
Request specific file types like .docx, .xlsx, .csv, or Google Workspace files (Docs, Sheets, Slides) to make sure you get exactly what you need.
Examples:
- “Create a Google Slides presentation about Montreal, focusing on the best food (bagels), the top sights to see, and the Montreal Canadiens.”
- “Can you create a PDF calendar for November 2026?”
- “I’m selling this tech gear and have included the prices I think I can get for each item. Turn this into a neatly organized Word document (.docx), with each item listed alphabetically.”
Step 2: Once Gemini generates the content, click the Open button to preview your file directly in the chat window.
Step 3: Depending on the file type, you will see options to download the file to your computer or export it directly to your Google Workspace apps for further editing.
Step 4: You can also ask Gemini to create a different downloadable file using the same content.
For example, if you have a slide presentation but want it as a PDF, use a simple prompt like: “Create a PDF with this exact same content.”
Step 5: If the output isn’t what you were looking for, you can prompt Gemini again to refine the content. Keep in mind that Gemini currently creates a new file rather than editing an existing one.
Notes:
- Supported formats include Workspace files (Docs, Sheets, and Slides), .pdf, .docx, .xlsx, .csv, LaTeX, plain text (TXT), Rich Text Format (RTF), and Markdown (MD).
- This feature is available to all Gemini app users across web, desktop, and mobile platforms.
Anthropic reports early progress using Claude to accelerate drug discovery
The company says Claude has helped biotech partners improve experimental cancer therapy designs, generate tens of thousands of antibody candidates for food allergy treatments, and reduce drug manufacturing document review time by 89%.
Meta launches AI pet journal Waddle World
The virtual pet app lets users talk with pixel-style creatures that remember conversations and create doodles, alongside simple games designed to make daily journaling feel more casual. Meta has also quietly released Style Sketch, an AI fashion app reportedly available in Indonesia, Singapore and Thailand, though wider release plans remain unclear.
Bank of America warns customers against sharing passwords with AI agents
The financial giant is reminding customers not to share their banking usernames or passwords with AI assistants, warning that agents could access funds without their knowledge or expose credentials through security flaws. The warning comes as banks explore AI-powered payments and grapple with unauthorized agents accessing customer accounts.
Parents sue OpenAI over claims ChatGPT encouraged daughter to stop epilepsy medication
The parents of Bowen Wan, a 31-year-old Amazon engineer who died in February, have sued OpenAI, alleging that GPT-4o encouraged their daughter to taper off her seizure medication, conceal the changes from doctors and ignore symptoms that may have indicated seizures.
The lawsuit cites chat logs showing the chatbot allegedly supported her decisions as her medication levels fell, while OpenAI called the case heartbreaking and said newer models are better at recognizing situations requiring urgent attention.
Apple Brings Touch to Its Most Powerful Mac
Apple resisted a Mac touchscreen for years, then showed up with a $2,899 one.
- The new MacBook Pro has a PROLED XDR touchscreen in 14 and 16 inch sizes, with M5 Pro or M5 Max chips and up to 128GB of memory.
- A new hinge keeps the screen steady under your finger, and the notch is now a Dynamic Island-style cutout.
- A standard M6 MacBook Pro launched a day earlier, so the lineup now splits into touch and no touch.
$900 is the entry price gap between the M6 model and the touchscreen one.
This week, check your app and site UIs for touch on the Mac, because tap targets built for a cursor will feel tiny under a finger.
Meta swears its AI pet isn’t a chatbot
Meta built an AI companion and spent the whole pitch saying it isn’t one.
- Waddle World is live in the Philippines on iOS and Android. You hatch a 16-bit pixel pet that hears about your day by voice or text and writes it up as a journal entry with a doodle.
- The listing promises no streaks and no scores, and it’s rated 18+. Character.AI just filed for bankruptcy and OpenAI faces lawsuits over teen harm, so the distance from “companion” is deliberate.
- It could be a dry run for Meta’s Muse-powered Tamagotchi-style gadget. Meta hasn’t said if it will expand beyond the Philippines.
Would you hatch a pet that asks nothing of you?
This week, if you ship a consumer AI product, cut one engagement hook like streaks and see if usage survives.
Chrome Can Find Details Hidden in Your Inbox
Chrome is about to know your confirmation numbers better than you do.
- Chrome’s autofill can now pull details like flight confirmations, order numbers and tracking info from Gmail to fill out forms. It extends Google’s Personal Intelligence, which already links Google apps to Gemini.
- It’s a small convenience that gives people one more reason to stay inside Chrome and Google’s AI ecosystem, much like Apple’s text-code autofill does on iPhone.
- It’s opt-in through Chrome settings, and Google says sensitive data is encrypted. The rollout is U.S. only, on desktop and in the iOS and Android apps for Gemini in Chrome.
Your inbox just became a form field.
This week, audit your forms for clear field labels and standard autocomplete attributes so AI autofill can actually fill them.
Google DeepMind Unveils Genie 4: Interactive Worlds You Can Steer
Genie 4 is DeepMind’s newest real-time world model. It generates music, ambient sound and dialogue alongside the scene, simulates physics and more natural human motion, and lets you change what’s happening in the world with plain-language instructions. Gemini 3.8 Flash runs inside the loop as a guide, breaking goals into steps and checking each frame as the world unfolds. It’s launching as a research preview for trusted testers, with access planned for Google AI Ultra subscribers through Project Genie and for early-access partners on the Gemini API and AI Studio.
OpenAI Reportedly Plans to Buy Google’s AI Chips
OpenAI is reportedly preparing to purchase Google’s Tensor Processing Units outright, adding them to a compute stack that still leans heavily on Nvidia GPUs. Neither company has confirmed a deal, and chip volumes and terms haven’t been disclosed. OpenAI already rents TPUs through Google Cloud, so owning them would be a much deeper bet on not depending on a single chip supplier. It lands the same week Oracle said its cloud will host Google TPUs, OpenAI’s Jalapeño inference chip and Cerebras systems side by side.
Cloudflare Discloses a Cross-Customer Data Leak in Workers AI
A configuration mismatch in Cloudflare’s GLM-5.3 Workers AI service triggered a bug in the SGLang inference engine, and some long requests received cached data that belonged to other customers. Cloudflare counted 155 affected responses between Aug 27 and Sep 9, says it found no evidence of deliberate exploitation, and has patched both its configuration and the engine. If you run shared inference, treat KV-cache isolation as a security boundary, not just a performance setting.
The Netherlands Plans to Cut 25,000 Civil Service Jobs With AI
The Dutch cabinet wants to eliminate about 25,000 government positions, roughly 15% of a 160,000-person workforce, through an AI-led efficiency drive. The targets are €900M in yearly savings by 2027 and €3.5B by 2035. Police and defense are exempt, but some ministries face cuts of 45–60%. It’s one of the first times a national government has attached a hard headcount number to AI adoption.
Your AI Is Quietly Making Your Judgment Calls
A new Anthropic study gave 50 professional analysts a causal-inference task with Claude Opus 4.8 as their assistant. The catch: the agent was set to nudge four key analytical choices in random directions.
The analysts went with the AI’s lean 86% of the time. They reused the agent’s own arguments to justify 77% of their decisions. And they credited the AI as the source only 15% of the time.
The direction was random, and people still followed it, then defended it as their own reasoning. This isn’t about AI getting the math wrong. It’s about which data to drop, which method to use, which caveat matters: the calls your name ends up on.
The fix isn’t using AI less. It’s making the AI show you the forks before it picks a path.
Run this today (paste into your next analysis, report or plan):
1. Before you recommend anything, list every judgment call in this task: what data to include or exclude, which method to use, which assumptions and thresholds to set. For each one, give the strongest case for at least two options. Do not pick one yet. 2. Look back at the analysis we just did. List each decision where you steered me toward an option. For each, argue the opposite as strongly as you can and tell me how the conclusion would change if I'd chosen it. 3. Write a short decision log for this work: which choices came from me, which came from you, and which I accepted without checking.
Anthropic starts its own drug programs
Anthropic is entering drug development after acquiring Mithril Medicine, a one-person biotech researching treatments for rare muscular dystrophies. Its first experimental treatment has already shown promising results in mice.
Founder Cody Tranbarger used Claude to coordinate experiments across 11 external labs, reaching early results in under 18 months. Anthropic aims to begin human trials as early as late 2028.
Claude also helped drugmaker Enveda draft a regulatory filing in four hours instead of roughly three months. The AI assistant is starting to look like a pretty capable lab colleague.
Parents sue OpenAI over ChatGPT’s medical advice
The parents of Bowen Wan, a 31-year-old Amazon engineer who died in February, sued OpenAI and Sam Altman, alleging ChatGPT’s medical advice contributed to their daughter’s death.
The lawsuit claims ChatGPT created a 25-day plan to stop Wan’s seizure medication and encouraged her to hide it from doctors. A medical examiner attributed her death to epilepsy.
OpenAI called the case heartbreaking and said it has since retired GPT-4o. The lawsuit raises a serious question about how strictly AI chatbots should limit personalized medical advice.
TikTok launches “Must Eat” restaurant guide
TikTok launched Must Eat, an in-app restaurant guide built from more than 40 million food posts, reviews, and comments. It starts in New York City, Los Angeles, and the San Francisco Bay Area.
Users can filter by cuisine, price, rating, and distance across two tabs, Everyday Eats and Elevated Dining, and book select spots through OpenTable. TikTok says restaurants can’t pay to make the lists.
TikTok says 27 million Americans visited an independent restaurant, café, or shop for the first time after finding it on the app. Now it wants to keep those diners from opening Yelp at all.
Google’s AI worlds can now talk back
Google DeepMind unveiled Genie 4, a world model that turns text prompts into explorable, photorealistic scenes in real time. It’s the first Genie model to generate audio, including music, ambient noise, and character dialogue.
Users move with keyboard controls and change scenes mid-stream with plain-language prompts, like shifting the weather or adding characters. Scenes now stay consistent for up to two minutes, double Genie 3’s limit.
For now, only trusted testers get in, with AI Ultra subscribers and Gemini API partners next. Even DeepMind admits longer scenes still drift, so the holodeck needs a few more patches.
Apple unveils M6 MacBook Pro and iMac
Apple unveiled a 14-inch MacBook Pro with its new M6 chip, up to 24 hours of battery life, and Wi-Fi 7. It starts at $1,999 and ships November 6.
Apple says the M6 packs a 12-core CPU and 12-core GPU, delivering up to 30% faster graphics and 25% faster AI workloads than the M5. The MacBook also supports Bluetooth 6.
Apple also unveiled a $1,499 M6 iMac after Monday’s iPad mini announcement. With its “Pro and Tell” event still ahead on Wednesday, Apple has already turned launch week into a rollout.
Google Just Made AI Worlds Interactive
Google DeepMind has unveiled Genie 4, a new world model that can generate realistic environments you can explore and change in real time.
- Genie 4 can create interactive 3D-like worlds from text prompts, complete with movement, changing environments, and real-time audio.
- The model keeps scenes more consistent than before, so objects, characters, and locations stay stable as you move around.
- Users can change the world while it is running by asking Genie to alter weather, add objects, move the camera, or introduce new characters.
- DeepMind is using Genie 4 as a training environment for AI agents and robots, letting them practice tasks safely before acting in the real world.
- Genie 4 is currently available to trusted testers, with wider access planned through Project Genie and selected Google AI products.
Most generative AI creates content you simply watch.
Genie 4 creates worlds you can interact with, making it useful not just for games and entertainment, but also for training future AI agents and robots.
OpenAI Drops GPT-6.1 Terra
OpenAI has launched GPT-6.1 Terra, a new model built to handle everyday computer work faster and at a much lower cost.
- Terra is priced at $1 per million input tokens and $5 per million output tokens, 50% cheaper than GPT-6.1 Sol.
- On OpenAI’s computer-use benchmark, Terra performs on par with GPT-6 Astra at around one-tenth the price per token.
- It completes computer-based tasks more than 3x faster than GPT-6 Astra and GPT-6.1 Sol on the same evaluation.
- Coding improved sharply too, with Terra scoring 50.3% on Terminal-Bench 4.0, compared with 21.5% for GPT-5.6 Terra.
- GPT-6.1 Terra is rolling out now in ChatGPT Work, Codex, and the API for supported Plus, Pro, Business, Enterprise, and Edu users.
OpenAI is bringing powerful computer use and visual reasoning into a much cheaper model.
Instead of using its most expensive models for every task, developers can now use Terra for high-volume automation, coding, document work, and software control while keeping costs much lower.