Episode
AWS News Mar 16 - Mar 22
- Podcast
- AWS News Weekly
- Published
- Mar 23, 2026
- Duration seconds
- 1017
- Processing state
not_requested
Actions
POST https://stenobird.com/v1/public/podcasts/aws-news-weekly-7066669/episodes/aws-news-mar-16-mar-22/transcription-requests
Idempotently request low-priority transcript generation for this episode.GET https://stenobird.com/podcast/aws-news-weekly-7066669/aws-news-mar-16-mar-22.md
Read the agent-friendly Markdown representation of this episode resource.
Summary
This week: Amazon Bedrock adds NVIDIA Nemotron, GLM 5, and Minimax M2.5 models for agentic AI workloads, while AgentCore Runtime gains shell command execution and WebRTC streaming. Amazon Redshift delivers up to 7x faster dashboard queries with new compilation optimization. Plus EKS Provisioned clusters get 99.99% SLA, Lambda functions can now discover their Availability Zone, and OpenSearch 3.5 brings persistent memory for AI agents. Connect & Contact Center Connect : Email forwarding to external addresses, 9 new TTS voices, 13 new locales, London for voice AI Connect Agents : Voice AI adapts tone/pacing to customer sentiment in real-time Data & Analytics SimpleDB : Export to S3 in JSON, cross-region/account, free tool Neptune : Query S3 directly from openCypher without loading Timestream InfluxDB 3 : Clusters up to 15 nodes, dynamic scaling Glue Data Catalog : IAM auth for S3 Tables and Iceberg views Redshift : 7x faster first-run queries, federated IAM Identity Center multi-region OpenSearch 3.5 : Persistent memory for agents, MCP integration, LLM-powered tuning Compute & Containers SageMaker HyperPod : Borrow idle GPU capacity beyond quotas SageMaker Training Plans : Extend GPU reservations 1-14 days Lambda : AZ discovery via metadata endpoint for zone-aware routing EC2 Fleet : Target interruptible Capacity Reservations in templates EKS Provisioned : 99.99% SLA, new 8XL tier doubles 4XL capacity ECR : Chainguard pull-through cache support AI/ML & Bedrock Bedrock AgentCore : Shell commands in sessions, WebRTC streaming Bedrock Models : NVIDIA Nemotron 3 Super, GLM 5, Minimax M2.5 Partner Central : Bedrock agents for co-selling automation NVIDIA NIXL : EFA support for disaggregated LLM inference Neuron DRA : K8s-native scheduling for Trainium Monitoring & Security Cl…