Today:

• OpenAI, Oracle, SoftBank Add 5 New Sites to $500B Stargate Project

• Google Labs Debuts Mixboard for Next-Gen Idea Mapping

• Alibaba Qwen Rolls Out AI Upgrades in Voice, Images, and Safety

• Sam Altman: OpenAI Aims to Build a Gigawatt of Compute per Week

• Microsoft Uses Agentic AI to Fast-Track System Upgrades

Plus, Wes Roth sits down with Ed Saatchi to get a glimpse into the future of how AI will disrupt Hollywood. “AI is not a tool. It’s a competitor”

OpenAI, Oracle, SoftBank Add 5 New Sites to $500B Stargate Project

OpenAI, Oracle, and SoftBank announced five new U.S. data center sites under the Stargate project, pushing them closer to their $500B, 10-gigawatt goal ahead of schedule. These centers will boost AI compute, create over 25,000 jobs, and support next-gen AI breakthroughs.

KEY POINTS

• Massive Investment: Stargate now exceeds $400B in committed infrastructure funding, on track to hit $500B and 10 gigawatts by end of 2025.
  

• Job Creation: Over 25,000 on-site jobs and many more indirect roles are expected from the new data centers.
  

• Nationwide Expansion: Sites were selected from 300+ proposals across 30+ states, highlighting U.S. focus and strategic partnerships.
  

Why it matters
This expansion supercharges America’s AI infrastructure. By scaling data centers quickly, OpenAI and its partners are making the powerful computing needed for AI widely available. It also brings real economic benefits—like new jobs and regional investment—while laying the foundation for future AI advancements.

Google Labs Debuts Mixboard for Next-Gen Idea Mapping

Google Labs has launched Mixboard, an AI-powered concept board that helps users visually brainstorm ideas using text prompts, images, and generative editing tools like Nano Banana. It allows easy project creation, image manipulation, and text generation, now available in U.S. public beta.

KEY POINTS

• AI Moodboarding Tool: Mixboard lets users create visual boards using text prompts, custom images, and AI-generated content.
  

• Nano Banana Editing: Google’s new image editing model enables natural-language image edits, image fusion, and regenerations.
  

• Public Beta Launch: Available in the U.S., with a Discord community for feedback and future updates.
  

Why it matters
Mixboard lowers the barrier to creative expression by turning ideas into visuals using AI. Whether for design, planning, or inspiration, it empowers non-designers to explore their imagination. It also marks another step in making advanced generative tools mainstream and playful.

Alibaba Qwen Rolls Out AI Upgrades in Voice, Images, and Safety

Alibaba’s Qwen AI team released updates across voice, image editing, and safety. Highlights include fast multilingual speech via Qwen3-TTS-Flash, improved facial and text editing with Qwen Image Edit 2509, and Qwen3Guard, a new moderation model that evaluates content in 119 languages in real time.

KEY POINTS

• Voice AI: Qwen3-TTS-Flash supports 10 languages, 17 voices, and delivers speech in just 97ms.
  

• Image Editing: Qwen Image Edit 2509 improves face, product, and text edits; supports multi-image fusion and control maps.
  

• Content Safety: Qwen3Guard moderates content in 119 languages with real-time and contextual evaluations.
  

Why it matters
Qwen’s new models show Alibaba’s strong push in global AI competition. They improve user tools in speech, visuals, and content safety—three essential areas for real-world AI use. These upgrades also help Alibaba compete with U.S. and Chinese AI leaders in both consumer and enterprise markets.

🧠RESEARCH

LIMI: Less is More for Agency

The LIMI model proves that powerful AI agents don’t need tons of data. With just 78 carefully chosen examples, it beats larger models trained on 10,000 samples. LIMI shows that smart demonstration—not size—unlocks real autonomy, helping AI agents think, act, and solve problems more efficiently in real-world tasks.

Qwen3-Omni Technical Report

Qwen3-Omni is a powerful AI model that handles text, images, audio, and video equally well. It beats top open- and closed-source models in 32 out of 36 audio-related tests. Using a smart architecture, it delivers fast, fluent, real-time responses—making it ideal for multilingual, multimodal tasks across speech, visuals, and more.

OmniInsert: Mask-Free Video Insertion of Any Reference via Diffusion Transformer Models

OmniInsert is a new AI tool that lets you insert people or objects into videos without using masks. It handles different subjects and scenes naturally, making the result look realistic. Using a smart training method and human-like preferences, it beats top commercial tools and works well even with limited data.

🛠️TOP TOOLS

Each listing includes a hands-on tutorial so you can get started right away, whether you’re a beginner or a pro.

What Will I Look Like in the Future Extrapolate - web app that uses AI to show how a face might age

Jungle AI – AI Flashcard Generator & Quiz Maker - turns your learning materials into interactive study questions

Harpy AI – AI Character Chat / Role‑Play - role‑play platform where you chat with AI characters

📲SOCIAL MEDIA

🗞️MORE NEWS

• Sam Altman outlines OpenAI’s vision to massively scale AI infrastructure, aiming to build a gigawatt of compute weekly. This expansion could unlock breakthroughs like curing cancer or universal tutoring—if compute bottlenecks are removed.

• Microsoft is using agentic AI to help companies modernize old systems faster. New tools in GitHub Copilot and Azure Migrate automate upgrades, simplify workflows, and reduce the time and cost of digital transformation.

• Former Spotify execs launched Oboe, an AI-powered platform that generates personalized learning courses from user prompts. While aiming to simplify self-education, it raises concerns about accuracy, as the content lacks citations or verified sources.

• By 2030, AI companies may face an $800 billion annual revenue shortfall, according to Bain & Co. Despite massive investments in data centers, revenue from AI services like ChatGPT lags behind soaring infrastructure costs.

• Microsoft has launched Gaming Copilot, an AI assistant for PC gamers that offers real-time tips and boss battle strategies. Available in Game Bar and soon on mobile, it uses voice mode and Bing-powered insights.

• UCLA and UC Davis are co-leading a $16M trial to test AI's effectiveness in breast cancer screening. The large-scale study will assess whether AI helps radiologists detect more cancers or causes unnecessary callbacks.

• Stanford researchers warn that AI-generated “workslop”—content that sounds polished but lacks real value—is quietly hurting workplace productivity. It clutters communication, wastes time, and erodes trust, prompting calls for better AI training and usage guardrails.

• The World Economic Forum’s new playbook outlines nine strategies for responsible AI adoption. It urges global businesses and governments to move beyond compliance, emphasizing trust, transparency, and cross-border collaboration as keys to sustainable innovation.