A developer newsletter covering several topics: an interview with Intrinsic CTO Brian Gerkey on AI-powered software-defined robotics and ROS; GitHub Copilot's shift to usage-based billing; Anthropic's potential Claude Code pricing changes; 12 agentic harness design patterns distilled from the Claude Code source leak; a Vercel OAuth security breach involving lateral movement through Google Workspace; and the ThoughtWorks Technology Radar's 'lethal trifecta' of agentic AI risk (untrusted content + private data + external communication).
Table of contents
1. GitHub moves Copilot to usage based billing2. My Cassandra complex is on fire3. The harness is the new product4. Can you vibe code an engineering intelligence platform? We find out.5. The 12 harness patterns you need to know6. A terrifying lateral security breach7. The lethal trifecta of agentic riskSort: