Find out why 1M+ professionals read Superhuman AI daily.

In 2 years you will be working for AI

Or an AI will be working for you

Here's how you can future-proof yourself:

1. Join the Superhuman AI newsletter – read by 1M+ people at top companies

2. Master AI tools, tutorials, and news in just 3 minutes a day

3. Become 10X more productive using AI

Join 1,000,000+ pros at companies like Google, Meta, and Amazon that are using AI to get ahead.

Sign up and start learning AI

Today:

• Meta Hires Three OpenAI Researchers

• Claude Now Supports AI App Building

• DeepMind Tackles Navier-Stokes With AI

• Suno Acquires WavTool for Music

• AI Replaces 50% Tasks at Salesforce

Meta Hires Three OpenAI Researchers

Meta has lured three top researchers—Lucas Beyer, Alexander Kolesnikov, and Xiaohua Zhai—from OpenAI’s Zurich lab to work on Mark Zuckerberg’s push for “super-intelligence,” meaning AI systems smarter than people. The trio set up Zurich for OpenAI after past work at Google DeepMind. Their move signals Meta’s determination to fix its shaky AI plans, add fresh brainpower, and keep pace in the race to build very advanced, general-purpose AI models soon.

Why this matters

1. Talent shift – Top minds leaving OpenAI for Meta changes where the newest ideas are born and tested.
   

2. Super-intelligence race – Meta’s hires show big tech is pouring effort into AI that could outthink people, raising both hopes and safety worries.
   

3. Competitive pressure – Faster progress at Meta will push rivals like Google, Microsoft, and Anthropic to innovate even quicker, speeding up the whole field.

Claude Now Supports AI App Building

Claude now lets anyone build, host, and share AI apps inside the Claude site. Describe what you want, and Claude writes code, fixes mistakes, and keeps the app running. Users sign in with their Claude account; their own plan covers the cost, so creators pay zero. Apps can show live React pages, read files, and chain several Claude calls. Current limits: no outside APIs, no saved data, text replies only.

Why this matters

1. Building AI tools becomes cheap and simple, inviting far more people to experiment.
   

2. Creators pay nothing when others use their apps, encouraging open sharing of AI games, tutors, and helpers.
   

3. Claude evolves from a chat bot into a full app platform, raising the stakes in the rivalry with OpenAI and other AI giants.

DeepMind Tackles Navier-Stokes With AI

Spanish mathematician Javier Gómez Serrano is working with Google DeepMind on cracking the Navier-Stokes puzzle, a 200-year-old set of formulas that describe how fluids move. Solving it wins $1 million and prestige. Their team uses AI to spot where fluid motion may explode, a key step toward proof. Gómez Serrano also helped train AlphaEvolve, a program that beats human mathematicians on tough problems, showing machines can speed discovery and reshape problem-solving.

Why this matters

1. AI as a math partner — Using machine learning to attack a century-old proof shows AI can handle deep theoretical work, not just pattern matching.
   

2. Tool for faster discovery — AlphaEvolve solving problems in a day hints at research cycles shrinking from months to hours.
   

3. Road to general intelligence — Success on Navier-Stokes supports claims that broad, self-learning systems could match top human experts this decade.

🧠RESEARCH

ShareGPT-4o-Image: Aligning Multimodal Models with GPT-4o-Level Image Generation

Researchers share a 91,000-item public set pairing words with images made by GPT-4o. Using it, they build Janus-4o, a model that converts descriptions or combined word and picture prompts into new, lifelike images after only six hours on eight high-powered graphics chips, inviting open exploration and learning for everyone worldwide.

Inverse-and-Edit: Effective and Fast Image Editing by Cycle Consistency Models

Researchers created Inverse-and-Edit, a method that turns any picture back into its generative blueprint, tweaks it, then rebuilds the result in only four quick passes. A loop-checking trick keeps key details while allowing changes. It matches top editing quality yet runs far faster, and the code is freely available online.

Outlier-Safe Pre-Training for Robust 4-Bit Quantization of Large Language Models

Outlier-Safe Pre-Training stops huge spikes in a language model’s inner numbers before they form. It replaces usual training steps with leaner ones, spreads energy evenly, and redirects tall peaks from word look-ups. A 1.4-billion-parameter model then keeps accuracy after shrinking to 4-bit size, easing phone-level use with tiny extra compute.

🛠️TOP TOOLS

LogoAI - Creating professional logos, cohesive visual identities, and automating brand promotion.

BoredHumans - Offers over 100 free AI tools, designed to cater to a wide range of creative, entertainment, and productivity needs.

Duplichecker - Designed to ensure the originality and integrity of written content.

Mnml AI - AI-powered design assistant specifically designed for architects, interior designers, and related professionals.

Getimg - AI-driven image generation and editing tool designed to create high-quality visuals from text descriptions.

📲SOCIAL MEDIA

🗞️MORE NEWS

• Suno, an artificial-intelligence music platform, bought WavTool, a web music studio, adding pro editing and stem separation—technology that splits vocals and instruments. The deal lets songwriters and producers craft original tracks faster inside a browser.

• Salesforce Marc Benioff says artificial intelligence now handles 30-50% of tasks, freeing employees for work. Layoffs followed the change. He calls it a “digital labor revolution,” boasting 93% accuracy yet conceding perfect results are impossible.

• A US judge said Anthropic legally trained AI on millions of books because the use was “transformative” (meaning it changes the originals). Dave Lee warns this ruling risks stripping writers’ income and weakening copyright protection.

• Retell AI uses OpenAI’s GPT-4o to create automated phone helpers that answer calls, schedule appointments, and resolve issues. Companies cut call costs 80% and gain 24/7 service. Eleven-person team built system quickly, fueling revenue growth.

• Authors Kai Bird, Jia Tolentino and others sued Microsoft in a US court, saying it copied 200,000 online books to train its Megatron AI. They seek $150,000 per book and a court order stopping misuse.

• OpenAI’s Deep Research API lets developers automate complex research tasks using GPT-based agents. It plans, searches, and generates structured, citation-rich reports—ideal for market analysis, healthcare reviews, or strategic planning—without manual prompting or coding.

• Microsoft’s deal with OpenAI stops it from creating its own human-level AI until 2030. Microsoft is blocking OpenAI’s restructure unless that rule is scrapped, sparking legal threats, revenue disputes, and rising rivalry between the partners.

• Deepseek has paused release of its next AI model, R2, because US trade rules block key Nvidia chips, causing shortages in China. The CEO dislikes results. No launch date. May update kept older R1 competitive.