AI Daily 2026.11.13: Jensen Huang Resigns as NVIDIA CEO; Anthropic Launches IPO; Gemini 4 Hydrogen Released

The day's top story comes from NVIDIA: its founder and face has stepped down overnight, prompting comparisons to Apple's removal of Steve Jobs. Anthropic takes a major step toward public markets with its IPO filing, while Google DeepMind rolls out Gemini 4 Hydrogen to the entire stack, and products keep shipping across meetings, developer tooling and localization infrastructure.

Funding & Markets: Anthropic Files IPO Registration Statement with SEC

Anthropic officially announced the launch of its initial public offering (IPO) and published the entry point ipo.anthropic.com. The company has filed a registration statement and prospectus with the U.S. Securities and Exchange Commission (SEC). Securities may not be sold until the registration statement becomes effective, and the preliminary prospectus is available through Goldman Sachs and Morgan Stanley.

Model Releases: Google DeepMind Opens Gemini 4 Hydrogen Across Full Stack for Coding and Agent Workflows

Google DeepMind officially released Gemini 4 Hydrogen, positioned for fast execution of everyday coding and agent workflows. It is now open across the full stack — Gemini API, AI Studio, Gemini apps, and Google Search.

Product Updates: Claude Meetings beta, Opus 5.5 becomes default model; Google Smart Glasses enter pre-order

Anthropic launched Claude Meetings beta, allowing Claude to join video calls as a visible participant that can answer questions during meetings, co-create documents and whiteboards in real time, and auto-generate notes and action items. At the same time, Claude Opus 5.5 was made the default model across Pro, Max, Ultra and Team plans, recommended as the daily driver for research, analysis and writing tasks.

Josh Woodward (Google Labs VP) announced that Google Smart Glasses have officially opened for pre-order, with trial experience available at select Warby Parker, Gentle Monster and Google stores; shipments begin December 10.

Replit Agent added two new capabilities: purchasing supplies on behalf of enterprises via Stripe Link and Visa, and creating and launching Google Ads directly within Replit covering Search and YouTube channels. OpenRouter also shipped localized UI in seven languages including Simplified Chinese and Japanese, with the full retranslation costing approximately $460.

Industry News: NVIDIA Board Announces Jensen Huang Steps Down as President, CEO and Director

The NVIDIA board issued an official announcement: Jensen Huang is stepping down immediately as president, CEO and director. CFO Colette Kress has been appointed interim CEO pending selection of a permanent successor, and the board has initiated an internal and external candidate search. VC analyst Tomasz Tunguz published analysis on the potential impact on the AI industry. Lex Fridman echoed a sentiment raised by Rohan Paul, comparing the move to Apple ousting Steve Jobs and citing past board-founder breakups at Atari, Uber and OpenAI to argue that boards periodically rediscover that founders themselves are strategy. Meanwhile, Democratic lawmakers Rep. Richard Neal and Sen. Ron Wyden launched an investigation into how the Trump administration handled NVIDIA's China-related chip sales and wrote to Commerce Secretary Howard Lutnick demanding preservation of related records.

Research & Safety: OpenAI Trains GPT with Automated Red Teaming; Epoch AI Maps Huawei-Horinger Compute Gap

OpenAI published a paper in its Alignment Research Blog describing how automated red-teaming improves GPT's *fuzzy robustness* in high-risk domains. Epoch AI released an analysis of Huawei's Horinger data center scale, revealing gaps between Chinese and U.S. high-performance compute deployments. Separately, World Labs open-sourced its world model Atlas, supporting keyframe generation, video clips and 3D Gaussian splats.

Quick Hits

  • Anthropic published a Claude Tag guide for startups to help teams make better use of annotation features.
  • Grok Bot 1.0 released as a collaborative AI teammate for teams.
  • Thinking Machines open-sourced Inkling 2 Small model weights under Apache 2.0 license.
  • OpenRouter published a tutorial outlining four layers of protection against AI agents stuck in infinite loops from repeated tool failures.
  • Figma opened testing for 3D transforms; a brand team shared how they produced a launch video using it.

What Developers Can Do Next

  • Want to try the latest coding-focused model right away? Access Gemini 4 Hydrogen directly via Gemini API or AI Studio, and deploy it through Celedog.io for low-latency routing.
  • Evaluating models for research and analysis workloads? Set Claude Opus 5.5 as your default driver and use Celedog.io's model routing to switch flexibly across Pro, Max, Ultra and Team tiers.
  • Building multilingual product experiences? Learn from OpenRouter's localization workflow — the entire site was translated into seven languages for roughly $460, about 6% of typical machine-translation pricing — and integrate through Celedog.io agents.
  • Shipping AI agents and worried about infinite-loop risk? Follow OpenRouter's four-layer defense guide to configure stopWhen step limits and API key credit caps on Celedog.io, keeping tool calling under control.

Written by Celedog · Last updated November 17, 2026

Where to go next