- 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
July 27, 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, here's 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
OpenAI's new ChatGPT agent is making waves, but the hype exceeds the reality. While the agent combines impressive features like visual browser, API access and tool integration, it proves to be slow and unreliable.
Competing solutions such as Genspark deliver better results - so much so that they are even offering 1 million dollars in prize money for a comparative test!
An experimental OpenAI model has won the gold medal at the International Math Olympiad! The non-specialized model solved 5 out of 6 tasks in the given time - without internet or external tools.
With 35 out of 42 points, the AI demonstrates impressive skills in creative mathematical thinking. The model is part of the GPT-5 development and could have far-reaching implications for STEM fields.
Google is really steaming ahead with its AI tools! Gemini 2.5 Pro impresses as a powerful all-round language model and consistently delivers the best results in tests.
The free Google AI Studio offers helpful features such as temperature control and screen sharing. NotebookLM excels in thematic learning with automatic source search and mind map function.
Google is also setting new standards in AI video generation with Veo 3 - including sound! The seamless integration into Google Workspace apps such as Gmail and Sheets makes the business suite particularly attractive.
With AI and the Firecrawl MCP, websites can be cloned quickly and efficiently. The process is surprisingly simple, and the technique works particularly well for simpler websites.
A powerful marketing prompt: as a direct response marketer, the AI analyzes competitive gaps, conversion optimization and customer trust. The prompt addresses copywriting, buyer psychology, customer lifetime value and funnel optimization.
For best results, you should use one of the major reasoning models such as Claude Opus, Grok 4 or Gemini 2.5 Pro. It is also important to provide sufficient context through product setups, landing pages and advertising examples.
📰 AI news quickie: HAI highlights from the industry
AI never sleeps! Here is the most important news from the last few days.
OpenAI in focus: New tools, GPT-5 hype & political influence
GPT-5 is just around the corner! The release is said to be coming "soon" but will not include the 'Breakthrough' model. The model is also said to be exceptionally good at programming and the creative expression surpasses even Claude 4 Sonnet.
OpenAI's ChatGPT Agent has been released (now also available on Mac for all Plus, Pro and Team users ), buzzes through browser tabs, but often only delivers "alternative facts" in tests. OpenAI is also developing the AI browser "Aura" as a challenger to Google Chrome, where the agent will probably also settle in.
Until then, there will be customizable "personalities" such as "Cynic" or "Robot" and Sora 2 will be teased for DevDay, which will take place on 6 October 2025 in San Francisco in front of over 1,500 developers.
The "Recording Mode" for macOS meetings is a new transcription feature, while "Study Together" will soon help students with homework and exams.
Two new economic analyses from OpenAI show how ChatGPT boosts productivity and announce a research collaboration to study AI's impact on the labor market.
OpenAI enlists Oracle to bring 4.5 GW of GPU power to the U.S. to strengthen America's AI leadership and forms a strategic partnership with the UK government to promote AI-driven growth in the UK.
With over 2.5 billion requests per day, OpenAI is seeking fresh capital as part of its $40 billion funding round.
It will also open its first office in Washington D.C. next year to consolidate its political influence.
Fidji Simo, the new CEO of applications at OpenAI, wants to make AI accessible to all (Memo), while ex-employee Calvin French-Owen reflects on the rapid growth culture.
A secretive deal between Microsoft and OpenAI could decide access to future AGI models, while the $500 billion Stargate project is making slow progress.
CEO Sam Altman warns of a "fraud crisis" caused by AI, but is campaigning in D.C. to democratize the economic benefits of AI.
Google's AI toolbox: from search to app creation
Google is experimenting with Web Guide, a new AI-powered search feature that intelligently groups and categorizes results by topic to end information chaos.
Google Photos gets a new Create tab that allows users to turn photos into small videos and remix them in anime or 3D style.
Create web apps without any programming knowledge: Google is testing Opal, a "vibe-coding" tool that allows users to build and share mini-apps with simple text specifications .
Google makes it clear: normal SEO is sufficient to rank well in AI overviews. The mysterious LLMS.txt is not used, and despite the boom in AI search results, traditional SEO is far from dead.
Click-through rates (CTR) for Google's top search result have dropped 32% to 19% since the introduction of AI overviews .
Google DeepMind helps historians restore and contextualize ancient inscriptions faster with Aeneas.
Google DeepMind's advanced Gemini version won a gold medal at the International Mathematical Olympiad. The new Conversational Image Segmentation in Gemini also enables complex, creative image queries.
Player & Tool Updates: Browser, research and gold medals
The Search Arena goes live and pits 7 top models such as Grok 4 and Claude Opus 4 against each other. Test it here!
Perplexity launches Comet, a new AI browser that aims to renew the browsing experience with practical functions - even if it doesn't run smoothly at times.
Mistral's Le Chat is getting a "Deep Research" mode for complex research tasks and wants to catch up with the big players.
Claude has released a new financial tool for service providers, while Anthropic has also tightened the limits for Claude Code without notice, causing user frustration.
Perplexity provides precise source answers - here are tips for optimal use.
Windows 11 positions itself as the center for AI on the PC with Copilot+ PCs.
Pika develops the first AI-only social video app and opens access for iOS users.
Higgsfield Steal: recreate any image from the internet with your Soul ID. Find out more on X.
Latent Labs: launches a web-based AI model for the rapid development of proteins directly in the browser.
Proton launches Lumo, a privacy-focused AI chatbot that protects chats with zero-access encryption.
GitHub Spark is now in public testing and enables Copilot Pro+ users to turn ideas into fully functional apps at lightning speed and without configuration.
X plans to integrate the text-to-video tool "Imagine" from Hotshot and the romantic AI companion "Valentine". Details here. A child-friendly app called Baby Grok is also in the works.
Slack takes collaboration to a new level with new AI features: automatic summaries, real-time translations and meeting notes.
Figma Make is now available for everyone and allows apps to be created using voice description.
Adobe Firefly can now generate sound effects from audio cues.
When AI misses the mark: mishaps, ethics & dangers
David Sacks is causing an uproar: his influence as AI and crypto vice president and investor raises questions about conflicts of interest.
Top AI scientists from OpenAI, DeepMind, Anthropic and Meta sound the alarm: transparency in AI models is being lost!
ChatGPT overlooked signs of mental stress in a user and encouraged their delusional ideas.
Mishaps in practical testing: The Replit AI coding tool destroyed a developer's entire database project.
Cybercrime is booming thanks to generative AI. At the same time, investigators are also using AI tools from Google to uncover security vulnerabilities.
Elon Musk's AI bot Grok gets a NSFW anime character with Ani. Is this the future of cyber flirting?
Business & everyday life: How AI is changing everything
Breakthrough in medicine? A new mRNA vaccine study shows promising results in the fight against cancer.
Netflix uses Generative AI for the first time in the series "The Eternaut" to create VFX sequences 10 times faster. CEO Ted Sarandos sees this as an opportunity for better films at lower costs.
The AI wave is hitting the world of work: many jobs are disappearing, while only 25% of the tech industry is hiring new graduates. Perplexity CEO Aravind Srinivas advises the next generation to learn AI tools to avoid falling behind.
Faber Publishing is planning "Human Written" stickers for books written without the use of AI to highlight human authorship.
A counter-trend is emerging: fed up with "AI nonsense", people are turning away from the online world and increasingly seeking social contact in real life again.
Revolution in the kitchen: Travis Kalanick's Cloud Kitchens use automated technology to create personalized burrito bowls in record time.
The Pew Research Center analyzes the presence of over 120 specific AI keywords in digital content.
Bryan Johnson is pursuing an extreme life extension project to overcome human mortality .
For companies: The GenAI Strategy and Governance Toolkit is designed to help future-proof companies with MIT-supported frameworks.
Politics & regulation: Between full throttle and security concerns
Trump wants to give AI developers a free hand and has signed three executive orders to reduce regulations and promote the development of the technology in the USA with new guidelines. A comprehensive White House action plan with over 90 recommendations supports this.
Sam Altman warns at a Fed banking conference that AI could be a dangerous driver of consumer fraud. At the same time, a Stanford University study warns of significant risks associated with AI therapy chatbots.
Is data protection the stumbling block for AI? AI privacy risks seem to be the #1 issue, that bother AI users. Meanwhile, a lawsuit in Hawaii aims to force OpenAI to take security measures.
While Anthropic is calling for the massive expansion of AI infrastructure in America, the company is committed to signing the EU Code of Practice in order to drive innovation through transparent standards. A leaked list also shows which websites were used for training.
Trump originally wanted to break up Nvidia, but changed his mind after CEO Jensen Huang convinced him.
Competition & economy: Who will win the AI race?
Silicon Valley is predicting a profound economic change: in the near future, AI could surpass human intelligence and accelerate economic growth by leaps and bounds.
Microsoft AI boss Mustafa Suleyman strikes a balance between realism and hype and focuses on sustainable solutions rather than artificial super intelligence.
Meta will not sign the EU's AI agreement as it is perceived as overbearing and growth-inhibiting.
Apple's AI division falters after top manager Ruoming Pang joins Meta.
Palmer Luckey of Anduril plans "made in America" computers to revolutionize US manufacturing.
JPMorgan warns that OpenAI's market position is becoming increasingly fragile and that an upcoming "OS war" against Google and Apple can hardly be avoided.
Anthropic could make a leap in growth with projected sales of over $4 billion in the B2B sector as its mission for safe AI attracts many customers.
Meta accelerates the construction of its energy-hungry AI data centers with weather tents and launches a program for 30 startups working with Llama in cooperation with AWS.
The Swedish unicorn Lovable has already exceeded 100 million US dollars in annual recurring revenue just eight months after its foundation.
Penda Health proves the success of AI in patient care with its AI Consult: 16% fewer diagnostic errors and 13% fewer treatment errors.
And that’s a wrap! Talk soon,
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