The Cloud Pod | Weekly AI & Cloud News on AWS, Azure & GCP · Justin Brodley, Jonathan Baker, Ryan Lucas and Matt Kohn | Cloud Computing & AI News

335: EKS Network Policies: Now With More Layers Than Your Security Team's Org Chart

·51 min·3 clips

AWS uses Bedrock to analyze JVM thread dumps across ECS and EKS in seconds, not hours.

1. The Cloud Pod episode 335 centers on EKS network policies, AI model releases, and cloud security changes across AWS, Google Cloud, and Azure. 2. Justin, Jonathan, Ryan, and Matthew host the weekly show, and Ryan and Justin drive most of the cloud security and AI commentary. 3. The episode’s core question is how new cloud features and AI deals change security, cost control, and developer workflow. 4. Meta’s new frontier model, codenamed Avocado and SucceedLlama, is described as a possible proprietary release after delays in training, performance, and testing. 5. The hosts point to Meta’s $14.3 billion spending since June and its leadership changes after Llama 4’s poor reception in April. 6. Disney sends Google AI a cease-and-desist over alleged training on copyrighted content and generation of Disney-owned characters like The Lion King, Deadpool, and Star Wars. 7. The Disney-Google dispute is framed as one of the first major legal tests for a content owner with deep legal resources against a cloud AI provider. 8. OpenAI then gets a separate Disney deal, with Disney investing $1 billion and licensing 200 characters for the Sora video app over three years. 9. The hosts treat that licensing arrangement as a sign that OpenAI is pursuing content deals after lawsuits over unauthorized training data. 10. OpenAI also ships GPT image 1.5, which the hosts say is up to 4x faster than the prior version and has better text rendering. 11. GPT-5.2 appears in three variants, including instant, thinking, and pro, and the episode cites a 70.9 win rate against human experts on GDPVal. 12. The hosts note that GPT-5.2 claims 11x faster output generation, near-100% long-context accuracy up to 256,000 tokens, and stronger safety responses for mental health and self-harm scenarios. 13. AWS opens Cedar to the CNCF as a sandbox project, with the hosts emphasizing RBAC, ABAC, REBAC, formal verification, and vendor-neutral governance. 14. GuardDuty Extended Threat Detection is described detecting a coordinated crypto-mining campaign that used compromised IAM credentials, modified instance attributes, and public Lambda endpoints without authentication. 15. The hosts also describe EKS admin network policies and application network policies as a two-tier control model with domain-name filtering for outbound traffic in EKS Auto Mode. 16. AWS’s automated Java thread dump analysis combines Prometheus, Grafana, Lambda, and Amazon Bedrock to diagnose JVM issues across ECS and EKS in seconds rather than hours. 17. The episode’s tone is conversational and skeptical, with the hosts frequently cross-checking whether a feature is genuinely useful for security teams, developers, or FinOps workflows. 18. The format stays news-driven and discussion-heavy, with quick reactions, short examples, and repeated side comments about how cloud tools fit into real operations. 19. Listeners who manage AWS security, AI tooling, or Kubernetes policy will get the most value. 20. Listeners wanting deep theory or a single-topic interview may skip it.

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from The Cloud Pod | Weekly AI & Cloud News on AWS, Azure & GCP