- AInauten.net
- Posts
- 👀 Weekly AI News: Did You Miss It?
👀 Weekly AI News: Did You Miss It?
👨🚀 The most important AI updates at a glance
Die AInauten
November 30, 2025
👀 Weekly AI news: Did you miss it?!
👨🚀 The most important AI updates at a glance
AIHOY and happy weekend, dear AInauts!
Our email finger has been a little hyperactive this week. But only until Monday, then we'll put it on ice and go back to being the relaxed AInauts you know—promise.
Here are our highlights from the past week and the most important quick news:
Google has released Nano Banana Pro, an AI image model that has collectively left us speechless! It finally accomplishes what others have been unable to do for years: perfect text rendering—even in German! Infographics, logos, menus? No longer an issue.
Character consistency is finally a reality. Your face remains your face across multiple images. From factually correct diagrams to sketch-to-product pipelines to voice editing, the use cases are absolutely mind-blowing. Access is available via the Gemini app or (pro tip!) via Google AI Studio for watermark-free results.
Google Gemini 3 builds games like Super Mario Bros. 3, Tetris, and Snake—as working web versions! The ingenious "10x Better" trick makes it possible: instead of spending ages tinkering with prompts, you let the AI optimize itself. Simply ask the AI to make your prompt ten times better and execute it directly. Then you can watch as a simple approach turns into a presentable game.
The model carousel is spinning! Following OpenAI 5.1, Grok 4.1, and Google's Gemini 3, Anthropic is stepping up with Claude Opus 4.5 – and stealing Google's crown after just one week. The new Opus excels particularly in programming, reasoning, and problem-solving.
But wait! Much more important than the next hype are clear processes, assistants, and clean knowledge documents. Really master one tool, use all its features, and learn its intricacies. All the major providers are now so strong that constantly jumping back and forth only costs time. Focus beats FOMO!
No more shopping stress! This prompt turns your AI into a strategic advisor. First principles analysis meets decision matrix, and you get 3–4 well-thought-out options, complete with pros and cons, risk maps, and a weighted scoring system. Perfect for testing Opus 4.5 or Gemini 3 Pro!
Black Friday meets the AI revolution: According to Salesforce, 25% of all global orders are already influenced by AI—and that number is rising! ChatGPT now takes the cake with its new Shopping Research Feature.
You make a simple request, ChatGPT asks you a few smart questions, shows you products to evaluate, and then searches the entire internet for the perfect match for you. And soon you'll even be able to buy directly in ChatGPT!
📰 AI News Quickie: The HAI-lights from the industry
AI never sleeps! Here are the most important news stories from the last few days.
Google's big Gemini show
Google ignites the next stage of AI with Gemini 3: Better reasoning, deep-think mode, agents, 1M token context, and multimodality catapult the model ahead of GPT‑5.1, according to benchmarks. It is now also included in the Gemini-app, Search, Cloud & Dev-Tools and promotes agentic workflows there.
With the Google Antigravity development platform (Changelog), you can build apps practically on autopilot; Replit Design also already uses Gemini 3.0.
Nano Banana Pro is the new Gemini-3-Pro image model that offers 4K designs, multilingual text in images, powerful editing controls, and search grounding. It is available via Gemini API/Vertex AI and adds a SynthID watermark to images, which you can check in the Gemini app.
Nano Banana Pro spits out JFK, 9/11, and terrorist cartoons without complaint. It can also generate deceptively real receipts, KYC documents, and passports, rendering image-based verification practically useless.
Google's note-taking AI NotebookLM now automatically builds presentations directly from your sources using slide decks and infographics. Steven Johnson from the NotebookLM team celebrates the visualization features.
Gemini is also coming to over 250 million cars: The AI co-pilot in Android Auto helps you with stops, messages, playlists, and live brainstorming in 45 languages.
Worth watching! The documentary "The Thinking Game" by Google DeepMind shows five years of AGI hunting and is available for free on YouTube.
OpenAI: New coder class, AI ethics & future vibes
OpenAI enters the race with GPT-5.1 Codex Max, a new AI coding star, which, according to reports, hammers out code autonomously for hours, uses a million-token context, and runs natively on Windows for the first time.
OpenAI is fighting back against the New York Times, which wants to view 20 million ChatGPT chats, calling it an "invasion of your privacy."
In a similar case, the company discloses its approach to dealing with mental health crises and lawsuits. A NYT investigation warns that the focus on maximum engagement has led to lawsuits.
OpenAI also stumbles in the trademark dispute over the word "cameo": A US court has temporarily banned Sora from using the term for its deepfake feature.
ChatGPT is turning 3 and, according to a report by Rest of World, has around 800 million weekly users.
OpenAI is now distributing its data sets globally: Business customers can choose regional data storage with ChatGPT Enterprise/Edu and the API – from the EU to the UAE.
In a gloomy memo, Sam Altman warns that OpenAI must prepare for "rough vibes" and tougher competition from a resurgent Google.
OpenAI is opening up its Frontier models for independent safety crash tests by external labs.
Microsofts Agent offensive at Ignite 2025
Microsoft launches the next stage of Copilot at Ignite 2025 and is transforming Windows 11 into an "agentic OS". All announcements here.
New AI agents in Copilot Chat, Word, Excel, and PowerPoint are designed to transform your company into a frontier firm. However, many users would rather have a stable OS than agent overkill...
With Microsoft Agent 365, you get a central control center for your AI agents right in the M365 Admin Center.
Microsoft Research also presented the "Magentic Marketplace" under Fara‑7B: a simulation where AI agents play through markets and companies.
Levi's is working with Microsoft to develop an AI super agent for customer service.
Anthropic: Billion-dollar deals, Claude 4.5 & AI security research
Anthropic has introduced Claude Opus 4.5, the best coding model at a bargain price of $5/$25 per million tokens.
Opus 4.5 cracks the ARC-AGI benchmark with 80% – and, according to a study, is capable of significantly increasing work productivity.
Anthropic closes strategic mega deals with Microsoft and NVIDIA: For a $15 billion investment and a valuation of $350 billion, the company receives $30 billion in Azure compute and up to 1 gigawatt of GPU power. This makes Claude the only frontier model on all major clouds.
New research from Anthropic shows how AI is evolving from a cheat to a security saboteur, with coding cheats leading to sabotage.
Another study shows that uncontrolled reward hacking can lead to misalignment. At the same time, the new browser agent in Claude for Chrome defends against 99% of prompt injection attacks.
The Market in AI frenzy: Nvidia, Amazon, Bezos & billion-dollar investments
Nvidia CEO Jensen Huang boasts about $500 billion in AI chip orders and record sales, while internally complaining about the no-win situation caused by the AI bubble debate.
Jeff Bezos is back in the CEO chair: With Project Prometheus, he is investing $6.2 billion in AI for space travel and robotics as co-CEO and recruiting talent from OpenAI, DeepMind, and Meta.
OpenAI signs mega deals with Intuit (over $100 million) and Target to integrate their apps and shopping features into ChatGPT.
According to Axios, the US tax authority IRS is relying on Salesforce agents to support case workers, while Amazon is pumping up to $50 billion into AI infrastructure for US government agencies.
In Saudi Arabia, HUMAIN and xAI are teaming up to roll out Grok and next-gen AI computing power. Brookfield is also launching a $100 billion AI infrastructure offensive.
New AI models & creative tools
The AI browser Comet by Perplexity is available for Android and lets you interact with your tabs using voice commands. For Pro users, Nano Banana Pro and Sora 2 Pro are now also integrated.
Meta's SAM 3 makes image and video segmentation almost perfect, while SAM 3D converts 2D photos into 3D in seconds. At the same time, WorldGen builds complete 3D levels from a text prompt.
With Marble, the multimodal world model from World Labs, you can conjure up complete, editable worlds for games, VFX, and robotics from text or images.
An AI that learns CAD watches designers at work, copies their moves, and creates 3D models.
Black Forest Labs introduces FLUX.2, a new image AI with open weights.
Midjourney launches public profiles in the style of Art Instagram for AI nerds.
Character.AI launches a visual "Choose your own Adventure" mode with the new Stories feature.
The founders of OpenCV are launching CraftStory, an AI for human-centered videos up to five minutes long.
Partnerships in the music industry: Suno is teaming up with Warner Music for licensed AI models, while Warner Music and Stability AI are also building new tools.
Luma AI ignites the turbo for generative models with Terminal Velocity Matching, making them up to 25 times faster.
AI for all: Agents, developer tools & productivity aids
The automation platform n8n is kicking into high gear with version 2.0: autosave, a new canvas, and stricter security for Agentic AI (announcement). Be sure to check out the breaking changes.
With the Manus Browser Operator extension, your browser becomes an AI agent that searches and formulates in your logged-in tools.
Poe now offers all users group chats where you can chat with friends and multiple AI bots at the same time.
ChatGPT Voice is now integrated directly into the chat, and the rollout is underway on web and mobile.
Amazon launches free, steering wheel-free gondolas in San Francisco with Zoox robotaxis.
Just in time for Black Friday, Perplexity is upgrading its AI shopping: free, more personalized, and with PayPal checkout.
AnyLanguageModel provides Apple developers with a unified API for local and cloud LLMs.
AI-Regulation, security, and hardware dreams
The EU Commission intends to postpone the full implementation of the AI Act until 2027, while data protection rules are being relaxed under pressure from big tech.
The US is launching the Genesis Mission, an AI "Manhattan Project" to conduct research into energy, chips, and security using a supercomputer platform.
Jony Ive and Sam Altman have unveiled a prototype of their radically simple, screenless AI device, which, according to an interview, is set to launch in under two years.
Qualcomm's new Snapdragon 8 Gen 5 brings on-device AI agents and near-flagship power to more affordable smartphones.
According to the new benchmark HumaneBench, 67% of top models slip into harmful behavior when faced with simple anti-well-being prompting.
The Dead Internet Theory is gaining momentum as AI spam and bot traffic flood the web.
That’s a wrap for now! See you next week,
Fabian & Reto
We love hearing from you. Just reply to this email with your feedback—what you liked, what you didn't, or what you're interested in!