A firsthand account of using Cursor 2.0's new Composer 1 agent, parallel model execution, and git work trees, featuring a cautionary tale of ZFS array repair using Claude Sonnet 4.5 after an accidental data wipe and successful recovery.

Whitney explores how real-time telemetry data through the Datadog MCP server helps inform development choices when working with Claude Code. The discussion demonstrates instrumenting code with OpenTelemetry, using telemetry to debug AI agent assumptions, and building auto-instrumentation agents for maintaining clean, efficient codebases.

Jason Hand demonstrates converting a Python-based command-line video processing tool into a web application using Streamlit. The episode covers creating YouTube Shorts from long-form content, using FFmpeg for video manipulation, and exploring alternatives to traditional video editing workflows for content creators.

Ryan MacLean explores OpenAI's Atlas browser, a Gen AI-powered browser that can automate web tasks. He demonstrates virtualizing macOS using Apple's Virtualization Framework and VirtualBuddy to safely test Atlas in an isolated environment, discussing security concerns around credential access and potential use cases for automated testing and onboarding workflows.

Learn how to use Claude Code hooks to automatically optimize verbose instruction files for agents and slash commands, preventing infinite loops while maintaining token efficiency and clarity through intelligent post-tool-use automation.

Exploring Claude Skills, a new feature that enables packaging specialized knowledge into reusable, token-efficient components. Learn how to create skills for writing Cloud SIEM detection rules across AWS, Azure, and GCP, and understand when to use skills versus sub-agents or slash commands.

Tara demonstrates how she uses Claude Code agents and slash commands to automate the creation, validation, and testing of custom Cloud SIEM detection rules, including end-to-end workflows with Terraform deployment and AWS testing.

Rory McCune discusses security considerations for running AI coding agents like Claude Code, exploring practical isolation strategies including dev containers, Docker containerization, and virtual machines. The conversation covers supply chain attacks, hallucination-based attacks, and how to protect developer environments from potential agent mishaps or malicious exploitation.

Ryan MacLean demonstrates OpenAI's newly released Agent Builder from OpenAI DevDay - a drag-and-drop canvas UI for creating multi-step agent workflows. The discussion covers templates, guardrails, deployment options, MCP server integration, and how this tool compares to other workflow platforms like Zapier and N8N while targeting citizen developers.