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:

• First Look at Google AI Ultra

• OpenAI Acquires Jony Ive’s Io

• Mistral Launches Devstral SWE Agent

• Google Expands Ads In AI Overviews

• Amazon Tests AI Audio Summaries

Google's MASSIVE "AI Ultra" | Mariner, Astra, Veo 3, jules, Diffusion etc (FIRST LOOK)

Google’s I/O update centers on a pricey new “AI Ultra” plan. For $250 monthly, users get early access to Gemini 2.5 model, V3 video maker, Flow editor, and expanded NotebookLM helper. Cheaper “AI Pro” stays.

 New tools include Gemini Diffusion that writes working code fast, Jules bug-fix bot, and Gemini Live assistant that sees your camera or screen. Project Mariner gathers web facts into notes. Early tests show promise despite small bugs.

WATCH THE VIDEO ON YOUTUBE

OpenAI Acquires Jony Ive’s Io

OpenAI will acquire Io, Jony Ive's design-led hardware startup, to fuse cutting-edge AI with imaginative hardware. Ive, former Apple design chief, and Sam Altman have spent two years exploring new devices that make AI feel personal and joyful. The Io team’s deep hardware talent now joins OpenAI, while Ive and LoveFrom take broad creative roles. The goal: reimagine everyday computing so people interact naturally with thinking machines everywhere, for everyone.

Why this matter

1. It links world-class hardware design with leading AI research, pointing to devices that move beyond screens and keyboards.
   

2. Ive’s user-first philosophy could make AI tools more intuitive, widening adoption and trust.
   

3. OpenAI’s leap into consumer products signals a new competitive front, pushing the entire industry to rethink how humans live and work with AI.

Mistral Launches Devstral SWE Agent

Mistral AI and All Hands AI have launched Devstral, a 24-billion-parameter open model that acts as a full software engineer. It reads big code projects, finds fixes, and writes changes. Because it is small, it can run on a laptop. Devstral beats many larger models on GitHub bug tests, scoring 46.8%. It is free under Apache 2.0, offers cheap API access, and plugs into tools like OpenDevin for coding work.

Why this matters

1. Broader access – A strong agent model that runs on ordinary laptops lets more people build and use AI privately, without costly cloud hardware.
   

2. Efficiency breakthrough – Devstral outperforms bigger closed models, proving smaller, open systems can still lead on real coding tasks.
   

3. Ecosystem boost – Its permissive Apache 2.0 license and agent-ready design can speed up development of autonomous software tools across industry and research.

Google Expands Ads In AI Overviews

Google says AI-powered summaries called AI Overviews make people search more and ask bigger questions. In the U.S. and India they raise usage over 10 percent. From today, Search and Shopping ads inside Overviews appear on desktop, with global rollout later this year. Google is also testing ads in conversational AI Mode. Existing Performance Max, Shopping and broad-match Search campaigns will automatically appear in these new spaces for advertisers worldwide.

Why this matters

1. Monetizing generative search – Blending ads into AI Overviews and AI Mode shows a clear business model for large-scale generative search systems.
   

2. User behavior shift – Higher search volume proves people trust AI summaries, hinting that future web traffic will flow through AI layers first.
   

3. New optimization frontier – Automatically surfacing existing ad campaigns in AI contexts pushes marketers and AI researchers to build smarter, context-aware advertising tools.

🧠RESEARCH

Emerging Properties in Unified Multimodal Pretraining

BAGEL is a powerful open-source AI model trained on massive mixed data from text, images, video, and the web. It understands and generates across different media types. BAGEL shows strong performance in complex tasks like image editing and video prediction, beating other open models and enabling deeper multimodal research.

SageAttention3: Microscaling FP4 Attention for Inference and An Exploration of 8-Bit Training

SageAttention3 is a faster and more efficient way to run attention in AI models. It uses new FP4 hardware to boost speed by 5× and introduces 8-bit training to save resources. While fine-tuning stays accurate, pretraining with 8-bit is slower. The method works easily with existing models.

VisualQuality-R1: Reasoning-Induced Image Quality Assessment via Reinforcement Learning to Rank

VisualQuality-R1 is a new AI model that judges image quality without needing reference images. It uses reinforcement learning to reason about visual differences and assign scores, mimicking how humans compare images. It outperforms existing methods, provides rich quality feedback, and works well across many image tasks like enhancement and generation.

🛠️TOP TOOLS

Sider AI - AI-powered browser extension and multi-platform tool designed to enhance productivity and streamline digital tasks.

Samplette -Web-based platform designed to assist music producers and beatmakers in discovering random music samples from YouTube.

Kits AI - AI-powered platform designed to revolutionize music production and audio content creation.

Godmode AI -  Web-based platform that harnesses the power of generative AI agents, offering users access to advanced automation capabilities inspired by AutoGPT and BabyAGI. 

MetaVoice Studio  - AI-powered platform designed to create high-quality, customizable voice overs for content creators

📲SOCIAL MEDIA

🗞️MORE NEWS

• Amazon is testing a new AI-powered audio feature that reads out product summaries and reviews. It helps shoppers save time by highlighting key details in a short conversation-style format, available through the Amazon app.

• Shopify launched an AI tool that builds online stores from simple keywords, letting merchants create layouts with images and text without coding. It aims to save time, boost creativity, and simplify eCommerce setup.

• OpenAI upgraded its Responses API with native image generation, Code Interpreter, remote tool integration via MCP, better file search, and new enterprise features. These updates boost agent capabilities, security, and ease of use—without changing pricing.

• Google has launched SynthID Detector, a tool that helps identify AI-generated content made with its models like Gemini and Veo. It scans images, audio, video, or text for hidden watermarks and highlights the AI-generated parts. The tool is currently in early testing with broader rollout planned soon.

• Meta has launched "Llama for Startups," a program offering up to $6,000/month and expert support to help early-stage U.S. companies build generative AI apps using its Llama models. This move strengthens Meta’s open model ecosystem amid rising competition and recent setbacks, including delays and benchmark controversies.