Daily AI Briefing

Friday, November 20, 2026

The editor's 6 picks from the 120 stories AIBriefs curated that day.

Get this briefing in your inbox every morning

EventBusiness15 sources

Anthropic prices record $107.3B IPO, shares jump 17%

Anthropic priced 980M Class A shares at $83.00, raising $107.3B in the biggest IPO on record, breaking SpaceX's mark. Shares rose 17% above the IPO price as ANT began trading on Nasdaq on November 19, 2026.

LaunchAI Models13 sources

Google's Genie 4 world model launches in Project Genie and Gemini API

Genie 4 generates continuous 720p worlds at 24fps with synchronized native audio, and lets users type prompts mid-session to change weather, lighting, or characters. It's live for eligible Google AI Ultra subscribers and, via the Interactions API, for developers in AI Studio.

AnalysisPolicy1 source

GAO tests: ChatGPT, Claude, Gemini and Grok aided attack planning

Government investigators bypassed safeguards on all four chatbots with common jailbreaks, getting usable attack plans in under 12 hours. Two systems helped with improvised-explosive plans in crowded U.S. cities; two supplied information on attacking electrical infrastructure.

EventBusiness6 sources

Nvidia tightens AI financing as interim CEO Kress sets terms

Nvidia's loan guarantees now only take effect once a neocloud secures independent financing and installs Nvidia gear, interim CEO Colette Kress said. Private-company stakes hit $65.5B, up from $22B in January, with $118B in guarantees including a $105B OpenAI Ohio backstop.

AnalysisAI Agents4 sources

OpenAI: coding agents now finish 46% of 2-4 hour research tasks

OpenAI says agents independently complete 46% of research tasks that take humans 2-4 hours, up from 23%, while researchers' AI use has risen exponentially since July. The top 10% of OpenAI researchers consume $23K+ of tokens per day, and human involvement remains central.

LaunchDevelopers5 sources

Perplexity launches Orchestrator API for model routing and failover

One endpoint covers 45+ open-weight and proprietary models from OpenAI, Anthropic, Google and xAI, plus GLM, Kimi and DeepSeek, with unified billing. Failover keeps the model fixed and swaps the deployment — a failed Claude request can be served by AWS Bedrock instead.

Also that day

  1. LaunchGoogle DeepMind releases Gemini 4 Neon
  2. LaunchGoogle releases Gemini 4 Carbon with agent 'swarm' feature
  3. EventJensen Huang resigns as Nvidia CEO after 33 years
  4. LaunchGoogle releases Gemini 4 Hydrogen for fast coding and agentic work
  5. EventAnthropic files S-1 for Nasdaq IPO under ticker ANT
  6. LaunchGoogle DeepMind unveils Genie 4 world model with native audio
  7. LaunchGoogle launches Gemini 4 Carbon, a frontier model built for cyber defense
  8. EventAnthropic IPO: Amazon and Google to own nearly 30% combined
  9. LaunchAlibaba launches Qwen4-Max: 2.4T MoE, 1M context, closed weights
  10. LaunchCognition launches Devin Bots, agents that own outcomes
  11. LaunchNVIDIA unveils Nemotron for Science, Rubin HPC GPU at SC26
  12. LaunchGoogle makes Gemini Tasks free in six countries

104 more stories that day — open the live feed.

Daily brief

Get tomorrow's AI brief in your inbox