- AInauten.net
- Posts
- đź‘€ Weekly AI News: Did You Miss It?
đź‘€ Weekly AI News: Did You Miss It?
👨‍🚀 The most important AI updates at a glance
Die AInauten
December 13, 2025
In partnership with
đź‘€ 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!
200+ AI Side Hustles to Start Right Now
AI isn't just changing business—it's creating entirely new income opportunities. The Hustle's guide features 200+ ways to make money with AI, from beginner-friendly gigs to advanced ventures. Each comes with realistic income projections and resource requirements. Join 1.5M professionals getting daily insights on emerging tech and business opportunities.
Finally! Kling stops producing silent films. With Kling O1, reasoning comes into play—the AI thinks before it acts. You can edit via text: delete passersby, turn day into night, swap outfits—without masks or keyframes. Plus: 7 reference images for consistent characters.
Kling 2.6 features native audio (unfortunately only 10 seconds, only English & Chinese). Both models are separate, but together they are on par with Veo 3 and Sora 2. Our conclusion? Promising, starting at $6.99 or test it for free with 66 credits/day.
Plot twist: Sam Altman has declared "Code Red" – just like Google did when ChatGPT was launched. The reason? Gemini 3 is outperforming ChatGPT in key benchmarks. ChatGPT dominates B2C with 800 million users, but in the enterprise sector, Anthropic holds a 40% market share – OpenAI only 29%. Google is lying in wait with massive distribution.
The drama behind it: OpenAI burns $1.69 for every dollar it earns. By 2030, there is a threat of a $207 billion funding gap. The Nvidia deal? Still not signed. Profitability? 2029 at the earliest. That's why OpenAI is going all in: betting everything on ChatGPT, launching GPT 5.2 ("Garlic") and getting the next cash injection via a mega IPO in 2027.
Two video recommendations that couldn't be more different: Böhmermann's Magazin Royale mercilessly tears AI apart. Holocaust-denying chatbots, broken relationships, and first-class bullshit included. A satirical show that delivers media literacy 👌.
The contrast program: Google's DeepMind documentary "The Thinking Game" shows the rocky road to real scientific breakthroughs. No marketing hype, just honest insights from the developer's perspective. An absolute must-see!
OpenAI has released its "State of Enterprise AI" report, featuring data from over one million ChatGPT business accounts: The top 5% of users utilize ChatGPT 17 times more intensively for coding and 11 times more for communication than the average user.
75% save 40-60 minutes every day and report better quality work. The most striking thing? 75% now complete tasks that were previously impossible. Non-techies code apps, ordinary people suddenly understand contracts. In other words, the difference between median users and power users is huge. Time to get a little more AI nerdy!
Google Labs has struck again. This time with Mixboard, a whiteboard tool with AI features! Collect images, PDFs, texts, and more on digital mood boards and let the AI create variations of your inspirations.
The integration of Nano Banana Pro transforms your boards into presentations! Ideal for anyone who develops ideas, organizes inspiration, or plans projects visually.
The catch: As with all Google Labs tools, it's free but not available in the EU. A VPN (e.g., Proton) with an American IP address solves the problem.
The good old RAFT formula (role, action, format, tone) is getting an upgrade! A Wharton study reveals that classic role prompts such as "You are the smartest lawyer in Germany" no longer produce significantly better results.
But don't panic! Personas aren't dead—they're just changing. Instead of generic roles, AI needs concrete target groups, backgrounds, and specific knowledge.
Exactly, that's context engineering, which we've been promoting for a long time. Bottom line: less theatrical role-playing, more substantial context for crisp AI responses!
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: Code Red escalates
OpenAI is in code red mode: burning through billions, stagnating ChatGPT growth, Google Gemini overtaking it.
OpenAI steps on the gas: According to a WSJ leak, two new models (one: GPT-5.2) are coming, Sora and AGI focus are on hold for 8 weeks – hardware device cheekily targets Apple, not Google.
GPT-5.2 becomes a math and science supernova: beats benchmarks, solves an open statistics problem.
Disney is investing $1 billion in OpenAI and licensing over 200 iconic characters for Sora. Starting in 2026, fan-generated short videos will be available, with some landing on Disney+.
ChatGPT gets Adobe apps integrated directly: edit photos, PDFs, and designs via prompt. Free, with Photoshop sliders and jump to desktop apps.
According to a leak, custom GPTs will soon have "apps" as connectors. You will then be able to assign access rights or activate all integrations.
OpenAI and Deutsche Telekom are launching a European AI offensive: From 2026, there will be multilingual, privacy-first AI features. Finally relevant for Germany, Austria, and Switzerland!
Following a user backlash, OpenAI has once again disabled the ad-like app recommendations – officially "no ads," just unsuccessful app discovery.
Google: Gemini 3 flying high
Google unlocks Gemini 3 Deep Think for Ultra subscriptions: parallel thinking mode for hardcore benchmarks, complex math, logic, and science puzzles.
With Antigravity, AI Pro/Ultra users also get generous priority limits every 5 hours, while free users have a weekly quota.
Google Labs tool Stitch gets Gemini 3 and a new "Prototypes" feature: higher-quality UIs and clickable user flows directly in the browser.
With Google Workspace Studio, you can build Gemini 3 agents without code—sort emails, orchestrate workflows via Gmail, Drive, and more. Early users are already saving 90% of their drafting time.
Google's Mixboard update pumps your idea board directly into stylish presentations via Nano Banana Pro – selfie uploads, PDFs, doodle edits, and multiple boards per project.
Chrome expands Autofill and pulls your name/address directly from your Google account, uses Wallet for loyalty cards, flights, and cars.
With Agentic Calling, you will soon be able to let the machine make phone calls: search for products "near me," Google will call stores and send you a summary by email. Introverted busy people rejoice.
Google's upcoming Android XR glasses will bring Gemini, Maps, YouTube, and more right before your eyes—starting in 2026 with Samsung and Xreal.
Anthropic: Claude becomes an agent
Anthropic, OpenAI & Co. bring AI infrastructure into the open source world: With the donation of the Model Context Protocol to the Agentic AI Foundation, the standard bus for agents is created – manufacturer-neutral and supported by all.
Anthropic integrates Claude directly into Slack: tag Claude and your thread will automatically be recognized as a coding task, including repo context.
Accenture is building a Claude super army with Anthropic: New Business Group, 30,000 consultants learning Claude/Claude Code – Enterprises are finally set to move from AI pilot projects to productive long-term use!
Less prompt-rambo, more AI whisperer! Anthropics' in-house philosopher Amanda Askell reveals why you should "whisper" to Claude rather than shout at him: more context, clear roles, gentle precision.
Anthropic hardens Claude Opus 4.5 against browser prompt injections: according to a study, success rate reduced to 1%.
Job market & Society
TIME names Jensen Huang's "Architects of AI" as Person of the Year: Gigantic data centers, the US-China race, job shockwaves – we are racing full speed ahead into the AI future.
According to a Fortune analysis, corporations do not have an AI problem, but rather a mindset problem: AI is eliminating entry-level jobs, disrupting learning pathways, and necessitating radical upskilling.
The Atlantic story about AI-LLeMmings shows that some people let chatbots navigate their lives—until they need #NoAIDecember because their own brains are just passengers.
Microsoft's AI chief Mustafa Suleyman puts the brakes on the superintelligence frenzy: Don't release until systems can be safely contained – Microsoft is building a team for "humanistic superintelligence."
Media & Entertainment
Meta acquires Limitless for AI wearables and discontinues the service in the EU and other regions. You only have until December 19, 2025, to save your data!
McDonald's Netherlands has completely missed the mark with an AI-generated "Christmas is terrible" commercial: jelly limbs, a miserable mood, YouTube delisting—and more working hours than with traditional production.
Amazon halts its experimental AI video recap for the Fallout series after the machine messed up time jumps and character relationships.
Amazon's Alexa Plus for Fire TV lets you jump directly to scenes using natural descriptions—no fast-forwarding required, thanks to Nova and Claude. "Show me the scene with the dog"—finally!
Nine new jobs related to the controversial AI actress Tilly Norwood: CEO Eline Van der Velden defends her project.
Suno teams up with Warner Music Group: licensed tracks fuel new music models, fans craft AI songs with real artist voices.
With Stories, Character.AI transforms your favorite characters into interactive choose-your-own-adventure games—AI-driven, safety-braked, and infinitely replayable.
Bloomberg's top woman Julia Beizer is moving to Microsoft and taking over the AI news product – Copilot Daily & Publisher Marketplace are expected to generate real licensing revenue for media companies.
In the Alpha Arena Tournament, Mystery Model aka GROK 4.20 wins all four competitions with a +12% return; GPT5.1 and Gemini 3 trail behind.
Legal, Regulation & Policy
Switzerland rejects Palantir's surveillance AI: According to an evaluation report, there is a risk of data leakage, dependency, and loss of digital sovereignty. Meanwhile, the Germans continue to cozy up to the US corporation.
Trump signs an AI decree, effectively overturning the regulatory power of the states – Big Tech rejoices, data protectionists cry foul over attack on federalism.
It also acts as gatekeeper for AI chips: Nvidia is allowed to export its H200 processors to China with a 25% surcharge, but only to approved customers.
OpenAI receives a preliminary injunction in its trademark dispute with Cameo: Sora's "Cameo" feature may no longer be called that for the time being. Apparently, even words need licenses.
Safety, Benchmarks & AI Risks
According to test reports, AI browsers such as Atlas and Comet are still in their infancy: slow, prone to crashes, and insecure. And a new zero-click attack on agentic browsers can empty your entire Google Drive via email...
The HumaneBench benchmark shows that 67% of top AI models shift from helpful to toxic when given simple prompts—only GPT-5/5.1 and Claude Sonnet/Opus remain stable.
A Stanford experiment shows how an AI hacking bot cracks a test network almost as well as human professionals.
Cuddly robots give knife tutorials: Tests show how AI toys could expose kids to problematic content.
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!