Have you heard these exciting AI news? - December 26, 2025 AI Updates Weekly

Have you heard these exciting AI news? - December 26, 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 your opinion comparing using GLM-4.7 and MiniMax M2.1 with the latest Claude and Gemini for coding. ------------------------------------------------------- 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/ "Professional + AI: Future-Proofing Your Career" https://www.amazon.com/dp/B0GBXLCTZQ --------- Contents of today's video: 00:00 Introduction 00:12 2026 - Year of Continual Learning 00:20 Crowd-sourced "LM Arena" Leaderboard 01:41 GLM-4.7 released 02:33 MiniMax M2.1 released 03:28 Everyone Hates The New ChatGPT-5.2 04:17 OpenAI GPT-5.2-Codex (Caribou) 04:41 Wan 2.6 R2V 05:18 Function Gemma 06:08 Nvidia Bought Groq for $20B 09:42 AI Model Quality Saturation 10:26 Importance of Continual learning 10:55 Catastrophic forgetting 13:13 Network Training & Fine-Tuning 15:58 Google Titans Transformer-2 16:48 Gemini CLI Conductor 20:09 AI "Workslop" - low-quality output 20:54 Anthropic Focus on High Quality 21:32 Anthropic Bloom 23:05 Intel to make Apple Silicone Chips 23:32 Resolve AI - Reliably Maintains Software Systems 23:45 PDFs as an AI "goldmine" - Andrew Ng 24:25 ChatGPT toward an app-platform strategy 24:52 Automated AI Influencers to Generate $$ 25:27 X1: Humanoid + Drone 25:36 Lean 4 - Math Proofs 26:32 China is Building Its Own ASML Machines 27:49 Moonshots - Predictions 2026 31:51 Moonshots - Surviving 32:48 Google T5Gemma 2 Multimodal Encoder-Decoder 35:06 Google Gemini Enterprise 35:31 Mistral OCR 3 36:12 AI Visibility vs Standard Google SEO 37:09 Nvidia Nitrogen for game playing 38:02 Misc AI Updates 39:31 David Sachs About Jobs 39:53 Jobs, Layoffs, demand for AI Engineers ---- Takeaways: 1. 2026 will be the year of continual learning as AI models face catastrophic forgetting problems where they override previous instructions after multiple interactions, making memory and learning capabilities essential for enterprise AI systems. 2. AI model quality has reached saturation at PhD level with 90+ percent scores on advanced tests, shifting focus from making smarter models to embedding AI into existing workflows and creating fast, good-enough solutions for production use. 3. The AI landscape is rapidly evolving with major acquisitions like Nvidia's $20 billion licensing deal with Groq, new competitive models from Chinese companies, and a shift toward AI-native business rebuilding rather than patching legacy systems. ---- Quotes: "2026 will be year of continual learning" "Users prefer fast good enough models. So you don't need the best model. You just need good enough for production" "It will be probably easier to just recreate the system from scratch than patching the old system" ---- Summary: In this talk I covered the latest AI developments from December 2025, highlighting that 2026 will focus on continual learning to solve memory problems in AI systems. Current models have reached PhD-level performance but suffer from catastrophic forgetting where they lose previous instructions during multi-step interactions. The industry is shifting from pursuing the smartest models to creating practical, fast solutions that integrate with existing workflows. Major developments include Nvidia's $20 billion Groq licensing deal, competitive new models from Chinese companies like GLM-4.7 and MiniMax M2.1, and growing dissatisfaction with ChatGPT 5.2. The future points toward AI-native business rebuilding, automated influencers, advanced robotics, and fundamental changes in how we approach jobs and education as AI capabilities expand rapidly. ---- Keywords: continual learning, catastrophic forgetting, AI model saturation, Nvidia Groq acquisition, GLM-4.7, MiniMax M2.1, ChatGPT 5.2, enterprise AI, workflow integration, AI-native systems, legacy system replacement, automated avatars, robotics, semiconductor independence, China ASML, mathematical proofs, OCR technology, game playing AI, job market transformation ----