AI Daily · Nov 13, 2026 · Friday
BAD INFLUENCE
NVIDIA says Jensen Huang has resigned as CEO and names Colette Kress interim CEO
NVIDIA's board announced that Jensen Huang has resigned as President, CEO and board member, effective immediately. CFO Colette Kress will serve as interim CEO while remaining CFO, and the board is searching for a permanent successor. Harvey C. Jones was named the company's first-ever board chair.
Product launches & updates
Claude BlogOfficialAnthropic makes self-hosted Claude Managed Agents available in private beta
Anthropic says self-hosted Claude Managed Agents is now available in private beta, letting teams run the full service, including the harness, sandboxes, credentials, and session history, in their own cloud or on-premise environment. Model requests can go to Anthropic, Amazon Bedrock, Google's Agent Platform, Microsoft Foundry, or a custom gateway. Access is requested through an Anthropic account team.
Tips & takes
AWS Machine Learning BlogOfficialAWS shows how to connect multi-agent systems across organizations with A2A
AWS explains how a LangChain enterprise orchestrator can call a Strands Agents-based SaaS compliance agent over the A2A protocol across separate AWS accounts. Each side authenticates with its own Amazon Cognito user pool, and the two sides share no secrets. The post includes three Jupyter notebooks for deployment and an end-to-end test.
OpenRouter BlogOfficialOpenRouter explains its self-correcting localization system for its website
OpenRouter says it built a daily loop that finds changed website strings, translates them with GPT-6.1 Sol, checks them in code, and has a second call verify meaning against the English. Failed checks are retried with the reasons attached, and reader reports become glossary rules, style notes, or test cases. The post says a full retranslation of 16,918 strings into 7 languages costs about $460 with verification included.
PyTorch BlogOfficialEUROPA explains how it trained a 1T-parameter MoE on JUPITER with PyTorch
The EUROPA consortium says it raised training throughput on the JUPITER supercomputer from 90 to about 1,300 tokens/s per GPU across 384 GH200 GPUs. The post describes the changes behind the gain, including FlexShard placement for the Muon optimizer, activation-checkpoint offloading to the Grace CPU, and running FSDP's all-gather on the compute stream.
AWS Machine Learning BlogOfficialAWS shows how to enforce per-user access in an agent's knowledge base search
AWS explains how to limit an agent's searches on Amazon Bedrock Managed Knowledge Base to the documents each user may view. The design has a request interceptor inject userContext from the verified token, with Policy in AgentCore and a resource-based policy as secondary defenses, so the LLM does not set user identity.