DOWN TO BUSINESS

🏭 HPE & NVIDIA unveil turnkey ‘AI factory’ stack for enterprise deployment

At HPE Discover, HPE and NVIDIA launched a new AI factory portfolio designed to accelerate enterprise AI adoption at scale. The stack includes servers powered by NVIDIA’s latest generation of high-performance GPUs, a next-gen private cloud AI platform, and over 70 ready-to-run AI workloads ranging from fraud detection to agentic AI. KDDI, Accenture, and 26 new partners are already on board.

💼 Goldman Sachs rolls out AI assistant firmwide

Goldman Sachs has launched its in-house generative AI assistant across the entire company, making it available to employees in all departments. The GS AI Assistant taps into multiple LLMs to help with tasks like document summarization, content drafting, and data analysis. Customized for roles from developers to investment bankers, the assistant is already in use by thousands.

🥤 PepsiCo deploys Salesforce’s Agentforce to scale AI-powered field operations

PepsiCo is rolling out Salesforce’s Agentforce platform across its operations to integrate autonomous AI agents into daily workflows. The move aims to supercharge customer engagement, field service, and marketing by combining human expertise with real-time data and intelligent automation. With Agentforce, PepsiCo gains unified customer profiles, smarter inventory visibility, and automated promotions.

🏗️ SoftBank pitches $1T AI and robotics mega-hub in Arizona

SoftBank’s Masayoshi Son is reportedly floating plans for a $1 trillion AI and robotics industrial complex in Arizona codenamed Project Crystal Land. The proposed mega-hub aims to replicate China’s Shenzhen in the U.S., bringing advanced tech manufacturing stateside. Son is seeking partnerships with TSMC, Samsung, and others while lobbying government officials for tax incentives.

TRENDING TOPIC

🎬 AI-generated commercial airs during NBA Finals

An AI-generated commercial aired during the NBA Finals this week, courtesy of betting platform Kalshi and AI filmmaker PJ Accetturo. Made with Google’s Veo 3 and stitched together in CapCut, the ad features bizarre vignettes such as an alien chugging beer and a man in a cowboy hat carrying a chihuahua. It took one person 2–3 days and cost just $2,000, about 95% less than traditional ad production.

HOT MODELS

🤖 Google launches on-device robotics AI

Google DeepMind has launched a new version of its Vision-Language-Action model that runs directly on robots, with no internet connection needed. Optimized for latency-sensitive environments, it enables bi-arm robots to handle complex tasks like folding clothes or unzipping bags, follow natural language instructions, and adapt to new tasks with as few as 50 demonstrations.

🚀 Google expands Gemini 2.5 family

Google has officially rolled out stable versions of its Gemini 2.5 Flash and Pro models, now ready for production use. Alongside them, it's introducing Gemini 2.5 Flash-Lite, the fastest and most cost-efficient model in the lineup. Flash-Lite is designed for high-volume, low-latency tasks like translation and classification, while still offering smart reasoning, multimodal inputs, and 1 million-token context, allowing it to handle large documents, complex tasks, or long conversations.

🔄 Meta launches V-JEPA 2 to help AI understand the real world

Meta has unveiled a new open-source “world model” designed to teach AI how physical objects move and interact. Unlike traditional language models, V-JEPA 2 helps machines like robots and self-driving cars make sense of their 3D environment, like knowing a ball will fall off a table when it rolls. By reasoning in a simplified internal space, the model learns without relying on labeled data or video.

🎥 Midjourney launches V1 of its AI video model

Midjourney has officially released its first video model. Users can choose from automatic or manual motion prompts, with options for low or high movement, and extend videos up to 20 seconds. It also allows you to animate your own uploaded images. Currently web-only and priced around the same as upscaling, the launch makes AI video creation 25x cheaper than anything previously on the market.

PLOT TWIST

🧠 New AI model diagnoses brain tumors with 99% accuracy

Researchers at Charité – Universitätsmedizin Berlin, Europe's largest university hospital, have developed an AI model called crossNN that can diagnose over 170 tumor types, including brain tumors, with up to 99.1% accuracy using just DNA from a fluid sample. Instead of relying on risky biopsies, the model analyzes a tumor’s unique epigenetic fingerprint and compares it against a massive reference database.

📝 Wikipedia pulls plug on AI summaries after editors revolt

Wikipedia has scrapped its plan to test AI-generated article summaries after fierce backlash from its editor community. The Wikimedia Foundation had proposed showing AI-generated overviews to 10% of mobile users, but editors quickly condemned the idea as “yuck,” “horrifying,” and a threat to the platform’s credibility.

🎬 Runway’s AI Film Fest crowns a mind-bending winner

Runway’s AI Film Festival has exploded in popularity, growing from 300 submissions in 2022. The grand prize went to Total Pixel Space, described as a film that "contains every possible digital image, including films of your entire life, every life you never lived, and the lives of every creature or object that ever and never existed on or off the earth, from every possible angle." A film following a chicken sent to human prison from the bird's perspective also won a $15K prize at the festival.