Have you heard these exciting AI news? - November 14, 2025 AI Updates Weekly

Have you heard these exciting AI news? - November 14, 2025 AI Updates Weekly Presented by Lev Selector Slides on GDrive: https://drive.google.com/drive/folders/1dmj00iilSP-JYmriMUUaHCKz5X54e3UX?ths=true Slides on GitHub - https://github.com/lselector/seminar (click on pptx file, then on "raw" or download button on the right) ------------------------------------------------------- Please write in comments if you started using Agents to execute tasks by writing code. Do you use Claude or another provider to run Sandboxes for that? Which method of running code do you use? CLI shell commands? Running scripts (Python or Typescript)? Or running functions that use "skills" text files? Can you share any useful tips? ------------------------------------------------------- Need AI Consulting ? - Enterprise AI Solutions - https://EAIS.ai - Linkedin - https://www.linkedin.com/in/levselector - GitHub - https://github.com/lselector ------------------------------------------------------- Our books: "Artificial Intelligence In Business: Strategies That Work" https://www.amazon.com/dp/B0FKZ3364G/ "AI Product Manager: Building Tomorrow" https://www.amazon.com/dp/B0FMS9NFK9/ "How to Learn with AI?: A Practical Guide for Students, Professionals, and Self-Learners" https://www.amazon.com/dp/B0FWLNVQMJ/ --------- Contents of today's video: 00:00 Introduction: AI Weekly Updates 00:07 "Let Agents write code, let code execute the tasks" 00:38 Crowd-sourced "LM Arena" Leaderboard 01:40 World Labs - Marble World Model 03:07 Code execution with MCP 05:39 OpenAI "GPT-5.1 Instant & Thinking" 06:35 OpenAI "AI-powered jobs platform" in mid-2026 07:19 Google Nested Learning 08:36 Gemini & Anthropic support JSON Schema 10:24 World Model RAG 11:51 Google Nano Banana 2 (Pro) preview 12:35 Verdent AI coding agent 13:03 Analyze Data in Excel or GSheets 13:24 Yann LeCun is planning to leave Meta 14:46 Microsoft's Humanist Superintelligence Vision 15:23 Baidu’s ERNIE-4.5-VL-28B-A3B-Thinking 16:12 China’s most affordable AI coding agent 16:53 Lovable Nears 8M Users (up from 2.3M in July) 17:29 Meta ASR Model for 1,600+ Languages 17:54 The "Magnificent 7" 18:38 Microsoft: more GPUs than Electricity 18:57 Nvidia invest in startups that buy its GPUs 19:15 Language Models are Injective and Invertible 20:16 AI Projects - succeed or fail? 21:38 rtrvr.ai v2 22:05 Google Colab VSCode Extension 22:31 Langchain introduce Sandboxes for DeepAgents 23:14 Cursor raises $2.3 Bln at $29.3 Bln valuation 23:38 Emad Mostaque Interview 25:35 SymPy for Math instead of LLM 26:12 Which LLMs are available in Moscow 29:14 AI for Mental Health 30:13 AI make money in Small Niches 31:48 Become AI Orchestrator 33:09 Making Money During AI Gold Rush 34:06 Jobs, Layoffs, demand for AI Engineers 34:18 Conclusion: Thank you --------------------------------------------------------------- ## Main Takeaways 1. Shift from Agent to Code Execution: A new trend, demonstrated by Anthropic, involves letting AI agents write code to execute tasks instead of performing the task directly by the model. This approach, exemplified by their MCP paper, significantly reduces token usage (by up to 98.7%) and increases precision, reliability, and speed. Anthropic, Langchain, and other vendors are starting to provide Sandboxes for code execution by AI agents 2. Coding Agent Dominance and Advancement. - Claude is noted as absolutely dominating the leaderboard for coding. - China's most affordable AI coding agent from ByteDance. - Verdant AI agent claims to surpass Cursor and Warp. - The vibe coding platform Lovable has also seen a massive user increase, reaching 8 million users 3. Some studies show a high failure rate for transformative AI projects (95% fail to deliver positive ROI), while success is found in AI augmentation (using ready-to-use models to improve existing work). Due to massive layoffs and the potential for human cognitive labor to become "economically worthless", the advice is to become an AI Orchestrator directing AI systems and building AI-native skills to seize opportunities in the "AI gold rush", especially in small niches. --- ## Top 3 Quotes 1. The slogan of today is "let agents write code and let code execute the tasks." 2. So what fails is when companies are trying to transform the work of certain group department or part of their business and they fail because transformation is difficult. But the companies which succeed are those which use ready-to-use models to augment, to improve what they're doing. 3. Within the next three years ... cognitive labor becomes economically worthless ... as humans become the dumbest people in the team."