Today:

• Qwen3-Coder Sets Agentic Coding Standard

• OpenAI, Oracle to Build 4.5GW Hub

• Meta Snags Three DeepMind Experts

• Microsoft Hires 20 DeepMind Experts

• Perplexity Seeks Smartphone Browser Deals

QWEN 3 CODER just set the AI Industry on fire

Alibaba’s Qwen3-Coder is a groundbreaking open-source coding model that rivals top proprietary AI like Claude Sonnet 4 and outperforms GPT-4.1 and Kim K2. The flagship 480B-A35B-Instruct variant offers 256K native context (scalable to 1M) and excels at real-world coding tasks through large-scale reinforcement learning and long-horizon planning. 

Qwen Code, its open-source CLI, enables seamless developer integration, making open-source AI coding tools nearly on par with frontier labs.

WATCH THE VIDEO ON YOUTUBE

OpenAI, Oracle to Build 4.5GW Hub

OpenAI and Oracle are building 4.5 gigawatts of new Stargate data center capacity in the U.S., boosting AI infrastructure beyond 5 GW and powering over 2 million chips. The $500 billion project aims to create 100,000+ jobs, support next-generation AI research, and strengthen U.S. AI leadership. With SoftBank and Microsoft partnerships, Stargate is becoming a global platform driving economic growth, advanced AI capabilities, and national competitiveness.

Why This Matters

1. Massive Compute Scaling: The 4.5 GW expansion significantly increases the compute needed for frontier AI models.
   

2. Economic and Job Impact: Stargate’s construction creates over 100,000 jobs, reflecting AI’s real-world economic influence.
   

3. Strategic AI Leadership: By partnering with Oracle and SoftBank, OpenAI strengthens U.S. dominance in global AI infrastructure.

Meta Snags Three DeepMind Experts

Meta has hired three AI researchers from Google DeepMind — Tianhe Yu, Cosmo Du, and Weiyue Wang — as part of its aggressive AI talent drive. These researchers previously worked on Google’s Gemini model, which demonstrated gold-medal-level problem-solving skills at the International Math Olympiad. This brings the total number of AI experts Meta has poached from DeepMind to at least six, signaling a high-stakes competition for elite AI talent.

Why This Matters

1. Talent Wars: The transfer of top researchers underscores intense competition among tech giants to lead AI breakthroughs.
   

2. Gemini Expertise: Hiring talent from the Gemini team gives Meta access to cutting-edge reasoning and math problem-solving capabilities.
   

3. AI Leadership Shift: Meta’s continued recruitment from DeepMind could reshape the balance of AI innovation between leading companies.

Microsoft Hires 20 DeepMind Experts

Microsoft has hired over 20 AI experts from Google DeepMind, intensifying the AI talent war. Among the high-profile hires is Amar Subramanya, former engineering head of Google’s Gemini chatbot. DeepMind co-founder Mustafa Suleyman, now leading Microsoft’s consumer AI, has also recruited talent to develop advanced AI health systems and consumer products. This surge in poaching highlights escalating wages, fierce competition, and the push to lead in artificial intelligence breakthroughs.

Why This Matters

1. Talent Consolidation: Microsoft’s hiring spree strengthens its AI expertise, directly drawing from a top competitor.
   
   

2. Rising Salaries: The poaching reflects skyrocketing compensation for elite AI engineers, shaping industry dynamics.
   
   

3. Strategic Shifts: With Suleyman at Microsoft, the company’s focus on AI products, including healthcare systems, is accelerating.

🧠RESEARCH

GUI-G^2: Gaussian Reward Modeling for GUI Grounding

GUI-G² improves how AI understands and interacts with app interfaces. Instead of using simple yes-or-no rewards, it models interface targets as smooth Gaussian curves, guiding clicks more accurately. This approach boosts precision, adapts

MiroMind-M1: An Open-Source Advancement in Mathematical Reasoning via Context-Aware Multi-Stage Policy Optimization

MiroMind-M1 is an open-source reasoning model designed for advanced math problem-solving. Built on Qwen-2.5, it uses a two-stage process—curated supervised training and reinforcement learning with context-aware policy optimization. It matches or surpasses top open-source models on benchmarks like AIME24 and MATH, while releasing all data, models, and training details for full reproducibility.

The Invisible Leash: Why RLVR May Not Escape Its Origin

This paper examines the limits of Reinforcement Learning with Verifiable Rewards (RLVR). While RLVR improves accuracy by refining known solutions, it restricts exploration and rarely uncovers new ones. The study highlights an “invisible leash” effect, where RLVR narrows answer diversity, suggesting future methods need stronger exploration mechanisms.

🛠️TOP TOOLS

SEO GPT - AI-powered tool designed specifically for search engine optimization tasks. 

StockImg AI - AI-powered platform that revolutionizes visual content creation. 

Plazmapunk - AI-powered tool that transforms audio files into visually stunning music videos.

Boords - Designed to simplify and streamline the video production process. 

StudyX - AI-powered educational platform designed to provide comprehensive homework assistance and learning support for students.

📲SOCIAL MEDIA

🗞️MORE NEWS

• Perplexity, valued at $14bn, is negotiating with smartphone makers to pre-install its AI-powered Comet browser, aiming to challenge Chrome and Safari. Backed by Nvidia and major investors, Perplexity plans rapid expansion to millions of users next year.

• Amazon is acquiring Bee, a startup that makes a $50 AI-powered wristband capable of summarizing conversations and creating reminders. The deal, part of Amazon’s AI push, follows its overhaul of Alexa and expansion into AI hardware.

• OpenAI and Penda Health tested an AI clinical copilot, AI Consult, across 40,000 patient visits in Kenya. The tool reduced diagnostic errors by 16% and treatment errors by 13%, showing AI’s potential to improve real-world healthcare delivery.

• Elon Musk’s xAI asked employees to film facial expressions for project “Skippy,” training Grok to read human emotions. Some workers raised privacy concerns over data use, while xAI simultaneously launched lifelike avatars Ani and Rudi, sparking controversy.

• Anthropic’s new research shows that giving AI models more time to “think” can worsen accuracy. Longer reasoning caused Claude and GPT models to overfit or get distracted, reducing performance on tasks like counting, deduction, and regression.

• Sam Altman told the Federal Reserve that AI will erase entire job categories, citing customer support as already transformed by AI. He claimed ChatGPT diagnoses better than most doctors but warned of AI’s destructive potential, including fraud and cyberattacks.