Plus: Google's back in smart glasses, OpenAI turns ChatGPT into a checkout counter, and a two-month-old startup just hit a $4.5B valuation.
Sponsored by
Kling AI is ranked #1 in image-to-video generation and just got even better. Their new O1 model combines video creation and editing into one workflow, and the 2.6 update adds native audio generation so you get complete videos with voiceovers and sound effects in a single pass.
Try it freeThis week in AI brought some massive plot twists. Trump flipped the script on chip export policy, IBM dropped $11 billion on enterprise data infrastructure, and the DOJ busted a $160 million smuggling ring while approving exports on the same day.
Meanwhile, Google's back in the smart glasses game, OpenAI turned ChatGPT into a checkout counter, and a two-month-old startup just raised $475 million at a $4.5 billion valuation.
If last week felt like everyone showing their hand, this week is the aftermath: deals getting made, bets getting placed, and the lines between who's winning and losing getting blurrier.
On this day in 2024, Google announced Gemini 2.0 and Veo 2, their breakthrough video generation model. One year later, we're watching everyone scramble to catch up to the foundation models arms race while quietly building out AI infrastructure that might take a decade to pay off.
Let's dive in.
🚨 Trump Greenlights Nvidia H200 Sales to China with 25% U.S. Cut
In a surprise reversal, President Trump announced Monday that Nvidia can ship its H200 chips to "approved customers" in China, with the U.S. taking a 25% cut of sales. The decision could unlock billions in revenue for Nvidia, though Beijing may restrict purchases anyway. Chinese President Xi Jinping "responded positively," Trump wrote on Truth Social. Blackwell and Rubin chips aren't included in the deal.
💼 IBM Snaps Up Confluent for $11 Billion in Largest Deal in Years
IBM is acquiring data streaming platform Confluent for $31 per share in an all-cash deal, marking one of its biggest acquisitions ever. Confluent's real-time data infrastructure is critical for AI deployment, connecting and processing data across hybrid cloud environments. The deal follows IBM's HashiCorp and Apptio acquisitions as Arvind Krishna doubles down on AI infrastructure.
🔒 DOJ Busts $160 Million Nvidia Chip Smuggling Network
The same day Trump approved H200 exports, the Justice Department announced "Operation Gatekeeper" had dismantled a sophisticated smuggling network that trafficked millions in Nvidia chips to China. Workers relabeled H100 and H200 GPUs under the fictional brand "SANDKYAN" before shipping. A Houston company owner already pleaded guilty; two more are in custody.
👓 Google Announces AI Glasses Arriving 2026 to Challenge Meta
Google's getting back into smart glasses with Gemini-powered AI eyewear launching next year. One pair is audio-only for hands-free assistant access; the other includes an in-lens display for navigation and translations. Hardware partners include Samsung, Warby Parker, and Gentle Monster. It's Google's biggest push into wearables since the original Glass flopped in 2015.
🛒 OpenAI and Instacart Launch Full Grocery Shopping Inside ChatGPT
You can now go from "what should I make for dinner" to "groceries arriving in 30 minutes" without leaving ChatGPT. Instacart is the first grocery partner to offer an embedded app with Instant Checkout directly in the chat interface. Powered by the new Agentic Commerce Protocol, it connects to 1,800+ retailers across 100,000 stores.
💰 Two-Month-Old Chip Startup Raises $475M at $4.5B Valuation
Unconventional AI, founded by former Databricks AI chief Naveen Rao, raised a monster seed round from Andreessen Horowitz and Lightspeed. The company is building brain-inspired neuromorphic chips aiming for 1,000x better energy efficiency than current silicon. Jeff Bezos participated, and Rao put in $10M of his own money. Just two months old.
📈 Fal Triples Valuation to $4.5B with $140M Series D
Turkish-founded AI infrastructure startup Fal announced its third funding round this year, now valued at $4.5 billion. The platform hosts 600+ AI models for generating images, video, and audio. Sequoia led the round with participation from Nvidia's venture arm and Kleiner Perkins. Revenue has grown 300% in four months.
⚠️ Moody's Chief Economist Warns AI Debt Is "Mounting Threat to Financial System"
Mark Zandi flagged that the top 10 AI companies will issue more than $120 billion in bonds this year, dwarfing Y2K-era tech borrowing even after inflation. Unlike the dot-com bust where losses hit equity investors, this debt-fueled buildout could spill over to credit markets if AI stocks correct. "Borrowing by AI companies should be on the radar screen," he warned.
⚡ Microsoft's VibeVoice is an open-source frontier voice AI framework that generates expressive, long-form, multi-speaker conversational audio—think 90-minute podcasts with up to 4 distinct speakers. It uses continuous speech tokenizers at 7.5 Hz and includes built-in watermarking for AI provenance verification.
🤖 AI Engineering Hub provides in-depth tutorials on LLMs, RAGs, and real-world AI agent applications. It covers everything from simple RAG workflows with LlamaIndex and Ollama to advanced agentic RAG with web search fallback, GitHub repository ingestion, and video-based semantic search using the Model Context Protocol.
📓 Open Notebook is a privacy-focused, open-source alternative to Google's NotebookLM. It supports 16+ AI providers (OpenAI, Anthropic, Ollama, Google), generates professional multi-speaker podcasts, offers intelligent full-text and vector search, and handles PDFs, PowerPoint, YouTube videos, and links—all while keeping your research local and secure.
🚀 Anthropic Quickstarts is a collection of Python projects that help developers build applications with the Anthropic API. It includes ready-to-deploy templates for customer support agents, financial data analysts with interactive visualizations, and the computer use demo that lets Claude control a desktop computer.
Coming in Hot
Create, send, and eSign business documents 40% faster.
Build and monetize your own branded AI chatbot.
AI-powered pixel art generator for game developers.
AI workspace for marketing teams to collaborate.
HTML to PDF conversion API for developers.
AI agent for semi-autonomous recruitment.
Your private AI agentic browser.
B2B revenue intelligence for smarter pipeline growth.
AI-powered UGC video ads in minutes.
AI-powered recruitment platform for faster hiring.
Submit your AI tool to the #1 AI tools marketplace.
What it does: Respeecher is an advanced AI voice generator and voice changer specializing in text-to-speech (TTS) and speech-to-speech (STS) conversion. It excels in emotion transfer technology to create authentic, nuanced voice replications, including cloning, de-aging, and multilingual dubbing. Ideal for media, entertainment, gaming, e-learning, audiobooks, and advertising, it breaks language barriers with support for dozens of accents, dialects, and languages while ensuring ethical AI use through consent and privacy protocols.
Key features:
Pricing: Respeecher offers flexible pay-as-you-go credit packs (billed per TTS characters or STS minutes, with volume discounts up to 50%) and monthly/annual subscriptions (17% savings on yearly, first month 50% off). Enterprise plans are custom-tailored.
Recent updates: Expanded multilingual library and ethical protocols highlighted in 2024 projects like "The Brutalist" and "Emilia Pérez."
What it does: CodeRabbit is an AI-powered code review tool that automates pull request (PR) analysis, bug detection (95%+ catch rate), and developer workflows. It integrates with GitHub, GitLab, and IDEs like VS Code/Cursor for contextual reviews, one-click fixes, PR summaries, and agentic chats—freeing devs to ship faster while maintaining code quality across large codebases.
Key features:
Pricing: Four tiers with per-developer billing (only for PR creators; reassignable seats). 14-day Pro trial (no card needed); free forever for open-source.
Recent updates: Free IDE extensions for VS Code/Cursor (2025); enhanced agentic chat for Jira syncing.
What it does: Vimeo is an AI-powered video platform for creators, businesses, and teams, focusing on ad-free hosting, editing, and distribution in HD/4K. Its AI suite automates translation, SEO optimization, viewer Q&A, and interactive learning videos—streamlining localization, highlights, chapters, and global content delivery for marketing, events, and e-learning.
Key features:
Pricing: Monthly/annual plans (40% savings on annual); Free tier available. AI features unlock in Advanced+; Enterprise custom. Upload limits are weekly/monthly based on videos.
Recent updates: Enhanced AI translation for 2025 global teams; new interactive learning videos (Nov 2025)
🚀 Stanford's AI Robot Makes History on the International Space Station
Stanford researchers achieved a first: using machine learning to help NASA's Astrobee robot navigate the ISS 50 to 60% faster than traditional methods. The system uses a "warm start" technique, training on thousands of past path solutions to predict optimal routes. Astronaut Sunita Williams assisted with setup while the Stanford team controlled tests remotely through NASA's Johnson Space Center.
🧠 NeurIPS 2025 Reveals the "Artificial Hivemind Effect" in AI Models
At the world's largest AI conference, researchers presented findings that 70+ language models generate eerily similar responses, a phenomenon they call the "Artificial Hivemind effect." The conference drew a record 26,000 attendees (double from six years ago) and highlighted a growing tension: AI capabilities are advancing rapidly, but scientists still can't fully explain how these models actually work.
🤖 MIT Develops Safety Control System for Soft Robots
MIT CSAIL researchers created a mathematically grounded framework that allows soft robots to safely interact with people and delicate objects. The "contact-aware safety" system blends nonlinear control theory with real-time optimization, enabling flexible robotic arms to grip items like grapes without applying excessive force. This could enable robots to assist in caregiving and food handling.
🚗 NVIDIA Releases First Open AI Model for Autonomous Driving Research
At NeurIPS, NVIDIA unveiled Alpamayo-R1, the world's first open industry-scale reasoning vision language action model for autonomous vehicles. The model integrates chain-of-thought AI reasoning with path planning, giving self-driving cars "common sense" to handle tricky scenarios like unexpected pedestrians or lane closures. It's available now on GitHub and Hugging Face for researchers.
🔬 NeurIPS Best Paper: 1000-Layer Networks Unlock Reinforcement Learning Breakthrough
While most RL models use just 2 to 5 layers, researchers built networks with up to 1,024 layers for robots learning goals without human guidance, achieving 2 to 50x better performance. The award-winning paper challenges the assumption that reinforcement learning can't effectively scale like language models, potentially opening the door to more capable robots and autonomous agents.
