Field Notes, 157 Articles
Notes from the foundry
Engineering essays on generative AI, retrieval systems, and what it takes to ship intelligent software to production.
21AI EngineeringThe Effort Dial Is an Architecture Decision, Not a Setting2026-08-14→22AI EngineeringThree Tiers, One Loop: Routing Between Model Sizes Without Breaking Your Agent2026-08-13→23Insights // InfrastructureKimi K3 Is Open Weight. That Does Not Mean You Can Run It.2026-08-12→24Insights // CostYour Agent's Real Unit Cost Is Per Completed Task, and You Are Probably Not Measuring It2026-08-11→25Insights // EvalsHow to Read a 2026 Model Announcement Without Being Sold To2026-08-10→26ComparisonsProvider SDK or Independent Framework: Pick by Where Your Control Flow Lives2026-08-09→27AI EngineeringMillion Token Windows Did Not Kill Retrieval. They Changed What It Is For.2026-08-08→28Tutorial // MCPCIMD Makes Your Authorization Server an HTTP Client. Plan For That.2026-08-07→29AI EngineeringLong-Horizon Agents Are Batch Jobs. Build Them Like It.2026-08-06→30AI EngineeringNative Multimodal Models Beat Your OCR Pipeline, Then Take Away the Thing You Needed Most2026-08-06→31Insights // DevelopmentWhy Trusting AI Generated Code Is the Wrong Goal2026-08-05→32AI EngineeringFive AI Agents, One Bug: When Missing Data Looks Like a Clean Result2026-08-05→33ComparisonsBest Open Weight LLMs for Agents in 20262026-08-04→34Tutorial // InfrastructureSelf-Hosted LLM Observability: Tracing Agents in Production with Langfuse2026-08-04→35Tutorial // AgentsWe Open-Sourced the AI Agent We Use to Scope Work and Watch Retainers2026-08-03→36Insights // TechnologyWhen Not to Build a Multi-Agent System2026-08-03→37Tutorial // SecuritySlopsquatting: How AI Hallucinated Packages Become a Supply Chain Attack2026-08-01→38Tutorial // SecurityAI Agent Security After the First Sandbox Escape2026-07-31→39Tutorial // MCPMCP Went Stateless: Migrating Your Server to the 2026-07-28 Spec2026-07-31→40Tutorial // DevToolsHow to Use Stacked Pull Requests on GitHub2026-07-31→