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

AWS Makes Kubernetes Conversational

·1 hr 28 min·5 clips

AWS Lambda managed instances lets you run Lambda on your own EC2 while AWS handles patching, scaling, and load balancing.

1. The Cloud Pod Episode 334, “AWS Makes Kubernetes Conversational,” focuses on AWS re:Invent 2025 and the week’s keynote announcements. 2. Hosts Justin, Jonathan, Ryan, and Matthew review predictions, and Matt Garman, Peter DeSantis, Swami, and Warner appear as the keynote figures that shape the episode. 3. The episode asks what AWS actually announced beyond the AI flood, and it uses the prediction game to frame the answer. 4. Jonathan’s pick for Lambda managed instances lands around AWS running Lambda on customer EC2, and the hosts connect it to serverless GPU support and existing EC2 spend. 5. Ryan’s wins include the new Graviton 5 instance family, the bigger model lineup in Bedrock, and a new Nova multimodal model. 6. Matthew’s best prediction is the new AWS Security Hub agent, which the hosts tie to automated SOC-style workflows. 7. The hosts also track the keynote count of “AI” and “artificial intelligence,” which totals 183 across Matt Garman, Peter DeSantis, Swami, and Warner. 8. Matt Garman’s keynote gets described as having better vocal delivery than before, but the hosts say the structure lacks a strong narrative thread. 9. The episode notes that Garman compresses 25 announcements into 10 minutes near the end of his keynote, which the hosts treat as a sharp contrast to the slower first hour. 10. Those 25 items include X8 instances, C8A instances, M8AZN, M3 and M4 Max instances, S3 vectors, OpenSearch GPU index work, and Security Hub general availability. 11. The discussion of AWS Lambda managed instances highlights run-your-own-EC2 semantics, AWS-managed patching and scaling, and support for Node, Java, .NET, and Python. 12. The hosts point out that managed instances let customers reuse EC2 reserved instances and compute savings plans while still writing Lambda-style code. 13. AWS Lambda Durable Functions becomes another major topic, and the episode explains it as sequential code with checkpointing, retries, and automatic state management. 14. The hosts note that Durable Functions can suspend execution for up to one year and restart from the last checkpoint after failures. 15. The episode also covers Bedrock’s new priority, standard, flex, and reserved service tiers, and the hosts react to the 99.5% SLA on the reserved option. 16. Nova Forge is discussed as a way for AWS to help customers build foundational models by blending customer data with curated training data across pre-training, mid-training, and post-training. 17. Peter DeSantis’ keynote earns praise for its infrastructure focus, including Nitro, Graviton, and the history-and-hardware style he uses on stage. 18. Warner’s keynote is described as the strongest narrative segment, with a time-machine video that argues development tools change but developers stay in control. 19. Engineers, cloud platform teams, and AWS re:Invent watchers who want concrete announcement details will get the most from this episode. 20. Listeners wanting a polished, low-chatter product-news roundup may skip the long keynote commentary.

As heard by us

AWS's durable workflow pitch gets a skeptical, operations-minded read.

This episode treats AWS's new durable state-machine feature as an operating decision, not just another keynote flourish. It draws a clean comparison with Azure Durable Functions, moves through JSON-defined states, and keeps returning to idempotency, monitoring, and the messy…

Read the full review in PlayNext →

Why you'd press play

You want the AWS feature that turns waiting into the whole problem.

Read the full recommendation in PlayNext →

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