Cord: Coordinating Trees of AI Agents lets agents build and run dynamic task trees with dependencies, parallelism, and human questions at runtime. It matters because outcome engineers need primitives for dynamic decomposition and context propagation to build reliable, observable multi-agent workflows — Principle 09 & 11.

Andrej Karpathy talks about “Claws” describes Claws: containerized, schedulable, message-driven local agents that persist workflows and run on-device. Running agent runtimes locally changes deployment, state management, and observability patterns that outcome engineers must design for when moving agents from experiments to infrastructure — Principle 07 & 09.

The Platform That Ate the Pipeline: Vast Data’s Rethink of AI Infrastructure reports Vast Data rearchitecting storage to deliver consistent, global, real-time data for always-on agentic AI, replacing batch pipelines. Outcome engineers building always-on agents must rethink data contracts, streaming semantics, and testability when pipelines are replaced by continuous, low-latency stores — Principle 04 & 06.

Making frontier cybersecurity capabilities available to defenders announces Anthropic’s Claude Code Security: an AI-powered scanner that finds complex vulnerabilities, verifies findings, and suggests human-reviewed patches. Treat model-driven vulnerability scanning as a first-class SDLC artifact and design human-in-the-loop gates, reproducible evidence, and audit trails for agent-driven fixes — Principle 14 & 03.

Interview with Notion CEO Ivan Zhao — custom Notion AI agents launching, agents build 50%+ of databases reports Notion launching custom agents that already build over half of Notion databases, embedding agents into core product workflows. That’s a concrete example of agents as product features; outcome engineers must instrument outputs, ownership, and lifecycle for agent-generated artifacts to maintain trust and traceability — Principle 09 & 03.