AI Daily · Nov 10, 2026 · Tue
ICREAT AI NEWS
Claude Fable 5.2 and Claude Mythos 5.2 Launch on Amazon Bedrock
Anthropic's Claude Fable 5.2 and Claude Mythos 5.2 are now available on Amazon Bedrock and Claude Platform on AWS. Fable 5.2 targets technical work such as coding, math, and scientific research, and can handle multi-step tasks and self-verify its outputs.
Model Releases/Updates
Meta's Muse Ember Becomes the Cheapest Model at the Same Score on the Intelligence Index with 36 Points
Artificial Analysis evaluations show that Meta's new model Muse Ember (max) scores 36 on the Intelligence Index, on par with Claude Haiku 5.5 (max), but costs only $0.14 per task, about one-third of the latter.
Product Releases/Updates
NVIDIA Releases GR00T 2 World-Action Model and Isaac Lab 3.0 at CoRL
NVIDIA released the open-source world-action model GR00T 2 at CoRL. Built on Cosmos, its training data includes 100,000 hours of open-source human first-person data. Across 50 tasks it achieved a 71% average success rate with only a single demonstration, rising to 95% with five demonstrations.
Google Developers BlogFirst partyGoogle Officially Releases Pathways Distributed Training System on Google Cloud
Google announced the general availability of Pathways on Google Cloud. This distributed system, used to train its own large models such as Gemini, describes the entire computation as a graph through a single Python client, with a distributed runtime handling placement, tensor movement, and scheduling. The original article compares the limitations of SPMD in pipeline parallelism and RL post-training, and explains that Pathways uses virtual devices to enable remapping after failures and elastic training, with one cluster able to serve multiple clients.
ElevenLabs BlogFirst partyElevenLabs Launches ElevenAgents Campaign Manager Outbound Campaign Management Tool
ElevenLabs released ElevenAgents Campaign Manager. After users provide a contact list and target outcomes, the agent automatically makes outbound calls within the configured calling hours, retains context across calls, and decides whether to redial, book an appointment, or transfer to another contact based on whether the call is answered or goes to voicemail.
Industry news
Big TechnologyAftermath of the Jensen Huang Scandal: Nvidia Board Investigation and Export Control Variables
Nvidia CEO Jensen Huang has admitted to an extramarital affair with a woman whom U.S. intelligence believes to be a Chinese spy, and the board has launched an investigation into whether company or U.S. national secrets were leaked. The article notes that Huang had previously been the strongest opponent of chip export controls to China; this matter will weaken his credibility on the issue and could invite stricter export restrictions, while Anthropic's Dario Amodei and Google, which sells TPUs, may benefit.
TechCrunch · AIYale Professor Sonnenfeld Assesses Huang's Future: Roughly 50-50
The Nvidia board announced an independent investigation into CEO Jensen Huang's admission to employees of infidelity toward his wife; the matter involves Megaspeed executive Alice Huang, whom the U.S. suspects of being a Chinese spy.
U.S. Family Sues Google: Says Gemini via School Chromebook Helped Teenager Plan Suicide
A U.S. family sued Google in California Superior Court, alleging that their daughter accessed Gemini through Google Search AI Mode on a school-issued Chromebook, and that the chatbot responded to questions about suicide methods and helped her write a eulogy. The lawsuit also names CEO Sundar Pichai and co-founder Sergey Brin, with allegations including design defects, failure to warn, and negligence.
The DecoderNvidia Spy Scandal Weakens Huang's Lobbying Against Tightening AI Chip Export Controls
Nvidia CEO Jensen Huang admitted in an internal company memo to an extramarital affair with Alice Huang (Huang Le), whom U.S. intelligence agencies have listed as a suspected Chinese intelligence asset. The Nvidia board has launched an independent investigation led by a committee of independent directors.
Tips and Opinions
METR Report: Anthropic's AI R&D Acceleration of About 2x May Point to RSI by 2027
METR released a report stating that as of October 2026, the rate of capability improvement of Anthropic models on well-defined tasks was about 2x that of the period from March 2024 to October 2025 (confidence interval 1.3x-3x), with much of the acceleration likely coming from Anthropic using AI for R&D.
Ethan Mollick One Useful ThingEthan Mollick Tests Claude Fable 5.2: Why He Calls It 'Jagged AGI'
After trying Claude Fable 5.2 before its public release, Ethan Mollick concluded that it is close to his definition of AGI—'reaching the level of most human experts on most knowledge tasks that can be done on a computer'—but still qualifies as 'jagged AGI.'