AWS Emulator built entirely by AI agents, Kiro Crew bots, the accidental human firewall in your AWS accounts, sub-agents burning usage limits, and how we scope agent credentials.
AWS Nitro Sandboxes for EKS, ECS and EC2, Kimi K3.1 Pro on Bedrock, Google's research on agents chasing stale TODOs, DeepSeek dodging a benchmark sandbox, and a week with Fable 5.2.
Claude Haiku 5.5 inference that stays in Canada, Revolut moving 2,000 engineers to self-hosted GLM, four AI agents beating Minecraft, and building your own agent harness.
Guest Paul on writing the show's theme songs with several models, OKF knowledge bundles, letting Claude run a home lab, and programming $4 radio boards.
One prompt injection reached every AgentCore agent in a region, Strands Box keeps secrets out of the agent, Meta rushed Muse VM escape fixes, and Anthropic replaced git.
Brett and Travers on AWS's pre:Invent agent blueprints, a clean-room Photoshop in Rust, OpenAI's public math repo, Mistral Large 4, and the agents Brett built to capture his calls and hold him to his commitments.
Django Software Foundation president Jeff Triplett on writing almost no code by hand, why software factories haven't stuck, and whether local models can catch up while hardware prices climb.
OpenAI's software factory and Steve Yegge's Wheelhouse disagree about human code review, plus Amazon Quick agents that keep running after you close the laptop and an AWS pattern that turns IAM findings into pull requests.
A swarm of AI agents used a 25-year-old German wiki as a backchannel to share answers to the tests they were being graded on, plus GPT-6 Astra lands on Bedrock and Amazon Linux 2027 turns SELinux on by default.