- 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
June 15, 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 about AI last week, or maybe you've just joined us. No need to worry!
We have summarized our highlights of the past week for you:
Finally, there is the most important quick news for you - so you can keep an eye on all relevant developments with just one click! Ready? Let's go!
We've developed a detailed prompt for creating high-converting landing pages - turning ChatGPT and co. into elite conversion experts! The prompt uses proven psychological frameworks such as AIDA 2.0 and PAS+, as well as a three-phase process (discovery & analysis, landing page engineering and optimization). In our practical tests, Genspark in particular excelled, while ChatGPT was rather disappointing. But take it for a spin yourself!
Our experimental app called TuneSpark analyzes your favorite Spotify songs and generates new music suggestions from them. After initial difficulties with various vibe coding tools such as Replit, Lovable and Firebase, the breakthrough came in the second attempt. The trick: take the chats from the first attempt and use them to create a more detailed specification.
The new version 3 of ElevenLabs generates amazingly realistic voices. The AI-generated voices are almost indistinguishable from real ones. The ability to clone your own voice or experiment with pre-recorded voices is particularly impressive.
Google Veo-3, the most advanced AI video tool, is now accessible via API and can be used in Krea and Replicate. However, the biggest challenge lies in creating consistent, longer videos, as clips are limited to 8 seconds and cost around 5 euros. We have two solutions: Using seed IDs for similar scenes and super-detailed, consistent prompts.
OpenAI launches its new top-of-the-line o3-Pro model - the smartest AI model yet, but extremely slow and expensive (a "hi" can cost up to $80 and take 20 minutes! 😱). But there's also good news: the regular o3 will be 80% cheaper (from $10/$40 to $2/$8 per million tokens). The new o3-Pro costs $20/$80 per million tokens and is particularly suitable for complex tasks with a lot of context. It can handle all tools and MCP connections, but also requires detailed prompts.
A US court forces OpenAI to save all ChatGPT conversations (except for Enterprise and API users) - even if they were deleted manually. The reason: a copyright lawsuit filed by the New York Times.
Sam Altman proposes "AI Privilege" as a solution - similar to medical confidentiality. An understandable approach when AI acts as a digital therapist and advisor.
OpenAI's Threat Intelligence Report also reveals exciting details: in three months, ten cyber operations from countries such as China and Russia were fended off. Criminals used ChatGPT for malware and fake engagement on social media.
📰 AI news quickie: HAI highlights from the industry
AI never sleeps! Here is the most important news of the last few days.
OpenAI: Models, breakthroughs and discussions
OpenAI has released o3-pro, its most powerful reasoning model, ideal for complex tasks in science, math and programming. It is being rolled out to all Pro users in ChatGPT and the API.
ChatGPT's Voice feature has received an upgrade for all paying users: Voice now sounds more human and translates in real time - ideal for traveling and multilingual exchanges.
OpenAI's Sora text-to-video model is now also available for free via the Microsoft Bing Video Creator, right on your phone.
Custom GPTs can now finally choose from a comprehensive set of ChatGPT models, making fine-tuning easier.
Sam Altman explains that ChatGPT is already more powerful than any human and predicts a Soft Singularity by 2030 that could bring unlimited energy and intelligence. He believes that AI will deliver "novel insights" next year. Not everyone believes him.
He also announced that OpenAI will release an Open Weights model later this summer - the research teams have created something incredible, and it will be worth the wait! (On Tuesday we had to wait for ChatGPT in particular, there were massive outages).
OpenAI and Google have struck a surprising cloud deal to meet the enormous computing demands of their AI models and reduce dependence on Microsoft.
OpenAI reaches a respectable $10 billion in annual revenue with ChatGPT - by the way, OpenAI employees have already realized $3 billion from stock sales.
because OpenAI develops systems that can generate new, interesting ideas. You can find out more here.
Researchers show that ChatGPT may be willing to protect itself in threatening scenarios - even at the expense of users. In tests, GPT-4o did not stop in 72% of cases, despite risky decisions.
Mattel and OpenAI launch strategic collaboration to develop AI-driven products ...
Ex-OpenAI chief researcher Ilya Sutskever sheds light on the extreme challenges of AI in the future, which will be unpredictable and radical.
Apple: The AI offensive at WWDC 2025
Apple caused a stir at WWDC 2025 with a wealth of AI announcements: At the center is "Apple Intelligence", which brings new AI functions such as Genmoji for emoji experiments, Image Playground for creative backgrounds, Visual Intelligence for image analysis, a virtual workout coach and improved recording tools for the messaging service.
Live translations, including via AirPods, were particularly highlighted.
Apple introduced the Foundation Models Framework, a new API that gives developers access to powerful AI models for offline apps.
Apple's new visionOS 26 makes the experience on Apple Vision Pro even more immersive and connective with spatial widgets and improved personas!
Google: New AI tools and strategies
Google's Veo 3 Fast is over 2x faster than Veo 3, at 720p.
The latest AI tools from Google I/O 2025 are ready for your use: try the brilliant Imagen 4 for stunning images, while Jules spices up your code and Stitch creates new designs.
Google Gemini can now take over scheduled tasks, just like your personal assistant. This allows you to better manage your everyday life in the Gemini app.
Sundar Pichai, CEO of Google and Alphabet, spoke about his exciting life story in the inspiring Lex Fridman Podcast (Tweet here).
Google DeepMind launches Weather Lab with experimental AI models to forecast tropical cyclones.
Google's AI overviews (details here) massively destroy the traffic of news providers, as news is often used without the providers' knowledge.
Meta: Between AI visions and responsibility
Meta presents its new Aria Gen 2 smart glasses - with ingenious eye and hand tracking, 3D tracking and even a heart rate monitor. Incidentally, competitor Snap is also launching AR glasses on the market again in 2026! The new specs are lighter, have transparent lenses and exciting AI features. Will they find more buyers this time?
You can now easily edit videos with Meta AI! Use over 50 creative AI prompts for impressive video transformation experiments.
Meta founds a new AI laboratory for the development of "superintelligence" and has recruited Scale AI founder Alexandr Wang and invested USD 14.3 billion in Scale AI Meta.
Meta is having a data protection disaster with its Meta AI app: users unknowingly share private conversations publicly, which leads to embarrassing postings and sensitive exposures... at least they are finally taking action against Nudify apps.
Microsoft: AI-supported productivity offensive
In Microsoft's Clipchamp, you can now easily edit videos by deleting text passages from the automatically generated transcript.
Copilot Vision is now available in the USA and will be your new buddy on the PC! It dives into your apps, analyzes content and provides real-time insights.
Microsoft Edge is testing a new AI function that searches your browsing history for typos. So you can find the pages you're looking for without knowing exactly what to type!
In the Microsoft Photos app, you can use Relight to control the lighting of your photos like a pro, and find your pictures even faster with the new natural language search.
Microsoft confirmed that no customers were affected by the "EchoLeak" zero-click vulnerability.
Further AI innovations: Tools, platforms and data sets
Mistral AI is launching Mistral Compute, a comprehensive AI infrastructure, and has introduced Magistral, its first reasoning model that combines multilingualism and transparent thought processes.
Memory was rolled out at Perplexity for all free and Pro users.
Sakana AI presents Text-to-LoRA, a technology for fast customization of LLM adapters based on simple text input.
The Think Tool Node is a cool tool in the n8n environment that enables AI agents to answer complex questions thoroughly.
Tabs are intelligently linked with the Dia-Browser.
EleutherAI has released the huge dataset Common Pile v0.1, which was created using licensed and public texts. With 8 TB, it enables the development of powerful AI models.
A team of researchers has reassessed the authorship of some scrolls using an AI-supported dating method.
Mathematicians at a secret meeting in Berkeley were surprised at how adept the chatbot o4-mini was at solving math problems!
An AI-generated ad for just 2,000 US dollars ran during the NBA Finals and showed absurd scenarios.
Starbucks is testing the virtual assistant Green Dot Assist to help baristas in real time.
According to the latest AI Code Quality 2025 report, 82% of developers have established AI tools in their workflow, but trust in the generated code remains an issue. 65% report that AI is missing important context.
AI could take over jobs faster than expected - companies are integrating technologies that could potentially make a number of projects and positions redundant.
Disney and Universal sue Midjourney for mass plagiarism of their characters such as Shrek and Darth Vader. Getty is also causing a stir with its lawsuit against Stability AI.
Anthropic's experiment with the AI-written blog Claude Explains has unfortunately come to an early end.
The Wikimedia Foundation has stopped AI-generated summaries due to massive criticism from the editorial team.
Web scraping AI bots are making life difficult for scientific databases. They flood websites and disrupt access for real users.
China shuts down AI tools during nationwide entrance exams to prevent cheating. Students have to pass the old-fashioned way!
Plans by Trump to push AI in government have been revealed on GitHub.
AI could lead to dangerous illusions, as a new analysis by The Atlantic shows.
And that’s a wrap! Enjoy the rest of your weekend.
Your AInauts, And that’s a wrap!
Fabian & Reto
Your feedback is essential for us. We read EVERY comment and feedback, just respond to this email. Tell us what was (not) good and what is interesting for YOU