Have you heard these exciting AI news? - October 24, 2025 AI Updates Weekly

Have you heard these exciting AI news? - October 24, 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/tree/master/2025 (click on pptx file, then on "raw" or download button on the right) ------------------------------------------------------- Please write in comments if you have experience or tips for using "skills" - Claude Agent Skills or local skills ------------------------------------------------------- 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 artificial intelligence weekly updates 00:13 “We're summoning ghosts, not building animals” - Andrej Karpathy 00:26 "Why Claude Skills is a Big Deal ?!" 00:33 Crowd-sourced "LM Arena" Leaderboard 00:49 Claude Agent Skills 05:17 Claude for Life Sciences 05:48 Claude Skills vs Agents 06:36 Replicating /skills Locally 07:38 Emulate Skills in Google 08:44 Claude Code for the Web 09:21 OpenAI Atlas Web Browser 10:39 Gemini "URL context grounding" via API 11:29 Vibe Coding Is Dead? 12:49 Run LLM on Android Phone 14:26 Cline CLI - run Multiple Instances 16:34 Preventing Loops in Multi-Agent System 18:32 Orchestration Frameworks 19:12 Deploying Agents 20:09 Veo 3.1 20:23 Nvidia Blackwell chip wafer - TSMC in US 20:39 Nvidia RTX Pro 5000 Blackwell 72GB GPU 21:15 ChatGPT with automatic memory management 21:26 Google Build - new "vibe coding" product 23:04 Unitree H2 Humanoid Robot 23:13 Google's Quantum Computing Breakthrough 23:38 Tencent Hunyuan World 1.1 23:54 Deepseek-OCR AI Model 25:34 Andrej Karpathy Interview 26:09 Google VISTA Video Agent 27:14 IBM Generalist Agent 28:39 Energy for AI 29:04 ChatGPT is losing to Anthropic, DeepSeek, ... 31:14 Four alternatives to JSON 32:12 Terminals on MacOS 33:11 Colleges in Trouble 34:21 Jobs, Layoffs, demand for AI Engineers *Takeaways:* AI Agent Development is Evolving with New Architectures: The introduction of Claude Skills is a significant shift, offering a method to codify expertise and make agent behavior predictable, transferable, and production-ready, in contrast to earlier multi-agent systems that primarily focused on simulating collaboration. The use of iterative self-improvement and iterative development is key in new agent systems like Google VISTA and IBM Cougar. ChatGPT's market share is declining, competitors like Anthropic Claude and Deepseek are seeing massive growth, particularly in the enterprise and coding markets. Claude is shown to be significantly more profitable per user and leads in coding workloads LLMs are increasingly being run on mobile devices using optimized frameworks like M&N Chat and Cactus to leverage NPUs. A new wave of vibe coding platforms, including the new Google Build, is creating huge competition in the no-code/low-code web development space. *Quotes:* "we're summoning ghost not building animals" (Andrej Karpathy) "So, Claude is much more profitable per user than OpenAI" "a country which will have energy for AI will win the race." (Peter Diamandis) *Summary:* In this talk, I provided the weekly artificial intelligence updates, focusing on the transformative impact of *Claude Skills* for building production-ready, expert agents by codifying expertise and using executable code. I then reviewed the highly competitive LLM landscape, noting that *ChatGPT is losing market share* in enterprise and coding to models like Anthropic and Deepseek, with Claude being significantly more profitable per user. The discussion also covered key developments in infrastructure, including running LLMs on *Android phones* and new *vibe coding platforms* like *Google Build* challenging existing services, as well as the new *Deepseek-OCR* model that achieves massive context compression by converting text to images. Finally, I touched on the shift in chip manufacturing and the declining value of a traditional college degree in the current job market. **Keywords:** Andrej Karpathy, Claude Skills, AI Agents, LM Arena, Gemini, Vibe Coding, Cline CLI, Multi-Agent System, Orchestration Frameworks, Veo, Nvidia Blackwell, ChatGPT Memory Management, Google Build, Unitree H2, Quantum Computing, Deepseek-OCR, Google VISTA, IBM Cougar, Energy for AI, JSON Alternatives, iTerm 2, Colleges in Trouble, AI Engineers.