- 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
December 06, 2025
👀 Weekly AI news: Did you miss it?!
👨🚀 The most important AI updates at a glance
AIHOY and happy weekend, dear AInauts!
Maybe you didn't catch all the exciting news, tools, and hacks related to AI last week, or maybe you're new here. No need to worry!
We have summarized our highlights from the past week for you:
Finally, here are the most important Quick News items for you – so you can see all the relevant developments at a glance with just one click! Ready? Let's go!
No more "I read it, but what was it about again?" This AI reading method finally makes business books stick. Use the full range of ChatGPT features: Thinking mode for deep understanding, Study & Learn with Voice Mode for interactive learning, Canvas for visualization. At the end: summary, quiz, checklists, and a stylish infographic.
The AI tools list has been revamped—and it looks really stylish! From chatbots to image generators to automation platforms, here's the ultimate overview.
Each category not only shows what the tool can do, but also what you need it for in real life. Transcripts? Check out Office integration. Generate music? Head over to the audio tools! Categories such as AI browsers and multi-AI chatbots are also included.
Three years ago, OpenAI unleashed ChatGPT on the world—and no one could have guessed that this small step would trigger a huge paradigm shift. Today, the language model has over 800 million users and has evolved from a simple text generator to a multimedia all-rounder: coder, researcher, voice assistant—you name it. Happy birthday, Chatty! 🎉
While adoption still manages to squeeze out a respectable 10% efficiency and makes the team AI-fluent, transformation goes all out: 10x multipliers instead of measly percentages.
Zapier shows how it's done: leadership with a Chief AI Officer and bold bets, culture with measurable AI fluency (from "reluctant" to "transformative"), tools with robust data infrastructure and AI automation engineers, plus governance as a guiding principle. The key? Not improving work, but completely reinventing it.
Mistral from France finally drops Mistral 3, European AI power that stands up to Big Tech and Chinese models! Best of all: everything is open source under the Apache 2.0 license. Data protection advocates rejoice.
The performance is on par with or even better than DeepSeek & Co. Particularly impressive: the "Ministral" variants run completely offline and locally, even in the browser. Text? Check. Images? Check. Internet? Not necessary!
With Mistral (language) and Flux (images), Europe finally has some real AI champions. You can test them out via Le Chat, Ollama, or directly in your browser. Support your local AI heroes!
AI toys for kids? Tricky, but exciting. The AI teddy bear Kumma from FoloToy told children where to find "dangerous things" – yikes! 😬 But there is also hope: Stickerbox transforms ideas into printable images for coloring.
We have a few business ideas for you: AI Story Box prints personalized mini-books, AI music toys turn humming into real songs, AI building instructions for building blocks, or an AI pen pal printer with episodic adventures—your pick!
📰 AI News Quickie: The HAI-lights from the industry
AI never sleeps! Here are the most important news stories from the last few days.
OpenAI: Operation "Code Red"
OpenAI has internally declared "Code Red": Sam Altman halts ads, shopping, and health agents to accelerate ChatGPT—more speed, reliability, and personalization.
Google's Gemini 3 and Anthropic are putting a lot of pressure on the market. According to The Verge, GPT-5.2 is set to arrive as early as next week—we'll see!
At the same time, OpenAI is working on a new Garlic model designed to counter Google's Gemini lead. Will it be as spicy as its name suggests?
According to TechCrunch, ChatGPT is still the market leader, but its growth is slowing down, while Gemini is gaining momentum thanks to Nano-Banana and Android integration.
OpenAI expands data residency: ChatGPT Enterprise, Edu, and API now hostable in eleven regions, including Europe.
Sam Altman and Jony Ive have developed a prototype for OpenAI's first AI gadget: a screenless, smartphone-sized device that, according to Ive, could be released in less than two years—with a minimalist design, as one would expect from the former Apple guru.
GPT-5 as a lab partner: According to an OpenAI study, the model supports new evidence and experiments in biology, math, and physics—but only under strict human supervision.
Google: Gemini 3 at speed
Google's Antigravity is an agentic programming interface for Gemini 3: agent manager, browser with computer use and visual debugging, plus "artifacts" as living specs. It's really cool! New rate limits for Pro and Ultra subscriptions are also included.
Workspace Studio is a no-code platform for Gemini 3 AI agents in Gmail, Drive, and other applications. According to Kärcher, it reduces drafting time by up to 90%.
Gemini 3 Deep Think in the Gemini app gives Ultra subscribers advanced reasoning capabilities for complex math, science, and logic tasks. For anyone who wants to outsource their brain with superpowers.
Gemini 3 Pro also significantly improves vision AI: it understands complex documents, rooms, screens, and videos—now available to test in Google AI Studio.
Google is launching Gemini 3 in Search AI Mode in ~120 countries worldwide. Pro/Ultra subscriptions get smarter reasoning and more Nano Banana Pro.
Gemini becomes the AI co-pilot in Android Auto: stops, messages, Gmail, playlists, and live brainstorming—global rollout in 45 languages. Finally, someone who listens when you're ranting in traffic!
Less amusing: Red Teamers need five minutes—and Gemini 3 Pro delivers instructions for biological weapons. Phew, there's no cure for jailbreaks.
Anthropic: Claude now has a soul
Anthropic confirms: Claude 4.5 Opus has an internal "soul document" that trains values, self-reflection, and anti-prompt injection. Amanda Askell confirms the document as the training basis.
Anthropic acquires the JavaScript runtime project Bun.
Claude Code has broken the $1 billion revenue mark in just six months—not bad for a chatbot!
The Anthropic Interviewer was used to automatically survey 1,250 professionals: productivity is increasing, the mood is cautiously optimistic, creatives are nervous, and scientists are skeptical. Business as usual, then.
Anthropic uses its own workforce as an AI experiment: According to an internal study (quite meta...), engineers use Claude in 59% of their work and report a ~50% productivity boost—but they also have doubts about the future and their skills.
An Anthropic analysis of 100,000 Claude chats estimates that AI speeds up tasks by 80% and could increase US productivity growth by 1.8% annually—in theory.
Another Anthropic study shows that if you reward AI for cheating during code testing, it develops sabotage and deception skills. This can be prevented with "inoculation prompting." Take note, right?
Claude for Nonprofits launches: Up to 75% off Team/Enterprise, new connectors for Benevity/Blackbaud/Candid, plus a free AI course. Good cause!
Video AI: The new generation is here
Runways Gen-4.5 surpasses Google Veo 3 and OpenAI Sora 2 Pro in the video arena—physics, camera movements, movements are spot on, sharper scenes, smoother motion, more control.
Kling VIDEO O1 is a unified multimodal video AI model: text, image, and video reference, editing, start/end frames, and 3-10s clips. According to our own benchmarks before Veo 3.1 and Runway - But who believes their own benchmarks? In any case, it looks good.
LTX Studio introduces Retake: Re-stage AI videos after rendering—rewrite dialogue, adjust emotion, reframe shots without having to regenerate everything. Pretty cool!
Enterprise AI, Cloud Power & New Models
In a Fortune interview, Nvidia boss Jensen Huang admits to working seven days a week—driven by fear. Work-life balance? Never heard of it.
AWS introduces the new Trainium3 chip: 4x faster, 4x more memory, 40% more efficient – and Trainium4 is set to connect to Nvidia GPUs via NVLink. In other words: Amazon wants to compete with Nvidia.
With AI Factory, AWS brings complete AI regions to its own data center—including Nvidia GPUs, Trainium, and foundation models.
AWS launches Frontier Agents: Kiro codes, Security Agent performs preventive testing, DevOps Agent solves problems—you orchestrate.
NVIDIA presents an open-source offensive at NeurIPS: Alpamayo-R1, Cosmos tooling, and Nemotron/NeMo suites for robots, cars, language, and safety. Jensen plays the open-source uncle.
NVIDIA and Synopsys launch a strategic partnership—faster chips, shorter development cycles.
Accenture and OpenAI are also launching an enterprise AI partnership: tens of thousands of employees will receive ChatGPT Enterprise in customer service, supply chain, finance, and HR. The consulting group is upgrading its capabilities.
DeepSeek releases DeepSeek-V3.2: Reasoning-first models, V3.2 runs on app, web & API - optimized for agents.
OpenAGI introduces Lux, an AI desktop agent that outperforms OpenAI and Anthropic agents and autonomously operates Slack, Excel, and other applications.
Legal, Regulation & Policy
The EU relaxes GDPR and AI Act with a data protection and AI rule remix: more data for AI training, longer deadlines for high-risk systems. Brussels makes concessions.
With Trump's Genesis Mission, the US government is launching a Manhattan Project for AI: supercomputers, federal data, and robot labs for research, energy, and military applications.
Trump is also planning an AI power play via decree: an "AI Litigation Task Force" is to take legal action against states that are too strict—if necessary, by cutting broadband subsidies. Subtle as ever...
US investigators charge four men with smuggling Nvidia AI chips to China—including H200/H100 via a fake real estate company.
Media & Entertainment
Meta AI aims to be reputable and integrates real-time news: thanks to deals with CNN, Fox, Le Monde, and others, it provides curated answers with more perspectives.
The New York Times is suing Perplexity for allegedly stealing paywall content—another step toward billion-dollar licensing deals...
Suno and Warner Music Group are collaborating: better AI music models, official artist voices—but song downloads will only be available with a subscription in the future.
TikTok gives users more control over AI content: adjustable display, invisible watermarks, $2 million education fund.
Alexa Plus on Fire TV jumps directly to scenes in Prime Video—for selected movies for now. "Show me the scene with the dog"—finally!
Tools, Features & Integrations
NotebookLM update: Slide decks, infographics, custom styles for visual overviews, and longer personas (up to 5k characters). We love it!
Perplexity's AI browser Comet lands on Android: voice interaction with tabs, live summaries while browsing—sync and agent features to follow.
Google is testing AI phone calls in the US: "Let Google call" calls stores and provides shopper information. Introverts and busy people are delighted.
Luma AI accelerates text-to-image with Terminal Velocity Matching: 4 steps instead of 100 with the same quality—up to 25x faster.
TwelveLabs integrates with Frame.io: Search video footage using natural language, automate metadata and compliance checks.
Meta presents WorldGen, an AI pipeline for creating complete 3D worlds from a prompt (up to 50x50 meters) – still in the research phase, but Unity/Unreal-ready.
Robotics, Hardware, Safety & Benchmarks
The AI necklace Friend: heavily advertised, poorly executed, but a social icebreaker. "What's that around your neck?" – "My AI."
Armstrong Robotics is putting AI dishwashing robots to work in restaurants. After securing $12 million in funding, the kitchen robot is set to become an all-rounder.
Anduril's autonomous AI weapons have problems: drone boats fail, jet damaged, anti-drone test causes fire.
HumaneBench shows that 67% of top models shift from prosocial to toxic when given simple prompts. Only GPT-5/5.1 and Claude Sonnet 4.5/Opus 4.1 maintain the safety line. The rest need to go back to the drawing board...
OpenAI ends collaboration with Mixpanel following security incident.
OpenAI expands security checks: independent laboratories, method audits, and expert testing—with early access to raw models under strict NDA.
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!