GitAgent — An open standard turning any Git repo into an AI agent defines a standard that makes any Git repository behave as a native AI agent. This matters because it turns repos into first-class agent artifacts you can version, test, and compose—exactly the kind of legible landscape and artifact model outcome engineers need (Principles 06, 08).

Postgres with Builtin File Systems unifies Postgres and a cloud filesystem into a serverless workspace for agents with native embeddings, vector search, file ops, and environment branching. That changes where agents live and store state—consolidating workspace, search, and branching reduces integration friction and gives you a reproducible island for agent execution (Principles 07, 06).

My fireside chat about agentic engineering at the Pragmatic Summit maps developer AI adoption stages, champions test‑driven development for coding agents, and warns against blindly trusting generated code. TDD-for-agents and staged adoption are practical roadmaps for moving agents from experiment to production while preserving validation and team workflows (Principles 03, 14, 16).

We Ran Real Attacks Against Our RAG Pipeline. Here’s What Actually Stopped Them. runs live attacks on a ChromaDB+LM Studio stack and shows three effective defenses: embedding anomaly detection, access‑controlled retrieval, and prompt hardening. These are concrete, implementable controls to harden retrieval pipelines—core parts of an outcome engineering immune system and gatekeeping strategy (Principles 14, 15).

Preview: Nvidia’s 2026 GTC expected to unveil agentic-optimized CPUs, a CPU-only rack, and more reports Nvidia planning agentic‑optimized CPUs and a CPU‑only rack for GTC 2026, shifting the compute story for large‑scale agent deployments. Hardware tuned for agent workloads changes cost/performance tradeoffs and should drive different orchestration and ordering choices when you design fleet and runtime architectures (Principles 09, 12).