- 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
May 24, 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!
The art of effective AI prompting separates beginners from experts. While many users treat AI like a simple search engine, true AI whisperers take a strategic approach: They design thoughtful conversations with clear roles, develop systematic prompt strategies and use model-specific techniques. Three essential guides provide support.
ChatGPT can do more than just generate text - it can also create a picture of your personal AI patron! It gets particularly exciting with the o3 model, which gives you an insight into the AI's "thought process". 🪄✨
Google Labs presents Stitch, a cool tool for creating app and web designs at lightning speed. After logging in with your Google account, you can choose between mobile apps and websites, with standard or experimental mode. The finished designs can be exported as code or Figma files. Themes and colors are customizable in standard mode.
Google's latest AI innovation Veo 3 makes even better videos - incorporating sound, music and dialog generation. However, the tool is currently only available with a US VPN and AI Pro/Ultra subscription in the Google Flow Editor. At the same time, Google is presenting Imagen 4 in response to OpenAI's recent image update. The new version offers 2k resolution, improved photorealism and more precise text display - perfect for comics and children's books. Imagen 4 is currently available for testing via Whisk (with a US VPN).
Do you know this? Sometimes ChatGPT's responses seem less thoughtful than expected ... is ChatGPT lazy because of high server load? Our "Thinking Prompt" asks the model to approach tasks more systematically: with multiple checks, critical analysis from different angles and special attention to possible weak points. Give it a try!
📰 AI news quickie: HAI highlights from the industry
What a crazy week - probably the craziest week this year. Here's the most important news from the last few days, let's dive in!
Google I/O 2025: Total AI integration
At Google I/O 2025, the focus was on over 100 AI innovations that overshadowed everything else (focus on AI here, more details).
The latest versions of the generative media models Veo 3 and Imagen 4 are state-of-the-art (more details here), and the new Flow tool is designed to help filmmakers with creative storytelling.
Google is also pushing ahead with Gemini as a universal AI assistance system. The Gemini app is becoming more personal and proactive, with features such as Gemini Live, Imagen 4 and Veo 3, and the Gemini 2.5 models are receiving updates such as "Deep Think". However, if you want to use everything, you have to book the new Google AI Ultra subscription model (for 249.99 $/month).
The new AI Mode in Google Search (details here) answers complex queries. In the shopping area, AI Mode will allow virtual try-ons.
Productivity will also be increased in Workspace: Google Workspace will be expanded with Gemini to include personalized email responses, meeting planning and real-time translations (more features for Gmail, Docs, Vids here). And NotebookLM now has video overviews and a mobile app for iOS/Android.
The autonomous code agent Jules is in public beta, and Gemma 3n is available for mobile AI.
Security & Accessibility: The SynthID Detector identifies AI-generated content. New accessibility updates for Android and Chrome include enhanced AI features for TalkBack, and the multilingual SignGemma model translates sign language.
Also cool: Project Starline becomes Google Beam, an AI-powered 3D video communication platform.
Google Cloud promotes AI competence at management level with a new, certified training course in generative AI, including free learning material.
Anthropic: Claude evolution with a focus on safety and developers
Anthropic presents the long-awaited next generation of its Claude models: Claude Opus 4 and Claude Sonnet 4, setting new standards. For Claude Opus 4, AI Safety Level 3 (ASL-3) safety standards have also been enabled to reduce abuse.
Developers benefit from four new features for agents on the API, including code execution and long prompt caching times of 1 hour. A bug bounty program has been launched for further security checks.
Microsoft's AI expansion at the Build Conference: from Copilot to Azure
Copilot integrations: Microsoft integrates ChatGPT's image generator into Copilot for Microsoft 365 for Word, Excel and Outlook. Outlook also gets new Copilot features for easier email management.
Windows & Edge AI features: Windows apps such as Paint, Snipping Tool and Notepad receive AI updates (e.g. AI sticker generator). Microsoft Edge introduces new AI APIs at Build 2025, including for PDF translation and page blocking.
Developer platforms & hardware: Microsoft presents the Microsoft Discovery agent platform to accelerate research, the Windows AI Foundry with new tools for AI developers, and highly efficient AI workstations for local training of AI models.
Azure and partner models: xAI's Grok 3 models are now integrated into the Azure AI Foundry, which could impact relationships with OpenAI.
OpenAI: Codex, API power and visionary plans
OpenAI has now integrated GPT-4.1 directly into ChatGPT. Codex, a new cloud-based software engineering agent that can handle numerous tasks simultaneously, has also been launched.
Developers can now use the Deep Research function to quickly analyze GitHub repos.
The Responses API gets support for remote MCP servers, image generation and code interpreters.
Future versions of ChatGPT could store and analyze your entire digital life, raising questions about privacy and ethics.
With the OpenAI to Z Challenge, participants can discover archaeological sites and win up to 250,000 dollars.
OpenAI is testing the integration of the Model Context Protocol (MCP) in ChatGPT, which makes third-party data accessible - a big step!
For more transparency in AI model testing, OpenAI has opened a new security assessment hub. In focus: Dealing with malicious content, jailbreak controls and hallucination ratings.
OpenAI is pursuing ambitious goals: the development of a superintelligence to solve global problems, while at the same time aiming to increase profits to 125 billion dollars by 2029. But there are also critics...
Sam Altman and Jony Ive present make a lot of hot air around io, a new, user-friendly product line of AI tools.
Meta: Between XR displays, llama promotion and AI model delays
Meta tests the first 3D Instagram photos on Meta Quest headsets, which are generated pixel-perfect from 2D photos.
Meta launches the Llama Startup Program to support startups with up to USD 6,000 per month in the development of generative AI solutions.
The release of Meta's latest AI model "Behemoth" has been postponed, triggering internal questions about the efficiency of the billion-dollar investment.
NVIDIA: Engine of global AI infrastructure and model development
NVIDIA CEO Jensen Huang predicts a multi-million dollar market for AI infrastructure.
In line with this, NVIDIA is introducing the Grace CPU C1 for edge, telco and storage, the new NVIDIA AI Data Platform for agentic AI and NVLink Fusion for semi-custom AI infrastructure.
In models and platforms, NVIDIA now offers cloud-to-robot compute platforms for humanoid robotics and the Llama Nemotron Nano 4B open reasoning model for scientific tasks.
Nvidia has succeeded in having US export restrictions on its AI chips lifted, while Huawei is subject to a comprehensive ban. Read more.
AI ecosystem: new models, developer tools and creative applications
Mistral launches the new AI model Devstral, which outperforms coding benchmarks and can be used commercially without restrictions.
GitHub introduces Copilot AI, a new coding agent that supports developers with AI-optimized tools.
Windsurf accelerates software development by 99% with the new SWE-1 models.
Stability AI introduces Stable Video 4D 2.0 for sharper, more consistent 4D asset production directly from single videos.
Runway launches the Gen-4 Image API for $0.08 per image, usable for virtual fitting rooms, games and creative tools.
ElevenLabs introduces the SB-1 Infinite Soundboard, an AI playground that generates sound effects based on text descriptions.
Xreal presents Project Aura, its second smart glasses for Android XR, which promises huge dual displays and hints at Gemini integration.
Acer presents AI-powered wearables: the FreeSense Ring for health monitoring and the AI TransBuds for real-time translation.
With Emergent AI, you can generate complete apps including backend and tests directly from chat inputs.
Have you heard of the facial recognition search engine PimEyes? It allows you to find your own photos on the internet - a double-edged sword between privacy protection and potential misuse.
AI is transforming industries: From e-commerce to healthcare
Klarna demonstrates increased efficiency through AI: revenue per employee rose to almost 1 million dollars, while expensive Salesforce contracts were terminated.
Shopify introduces an AI store builder that creates online stores with a single command; the AI assistant Sidekick now also provides support via voice chat.
Amazon Music tests the AI search function "Explore" to discover artist information and music highlights.
Salesforce integrates new digital teammates into Slack, specialized AI agents that are designed to perform complex tasks in chat and compete with Microsoft Copilot.
AI startup Cohere acquires market research platform Ottogrid to strengthen its enterprise tools for automated market analysis.
Netflix plans to use AI to seamlessly integrate commercials into content to match the respective program.
In Saudi Arabia, a clinic experiment was launched with the AI doctor "Dr. Hua", who analyzes symptoms and suggests treatments.
Perplexity is expanding its offering with new Pro Perks for subscribers, including deals from brands such as Visa and access to premium data sources.
Two AI pioneers warn of the existential risks of artificial superintelligence in their (soon to be published) book If Anyone Builds It, Everyone Dies.
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