The notebook / ideas in practice
AI.
Notes on agentic AI, building agents, and developing software with AI assistance.
Pipelines, Part 2: Drafts, Audit Trails, and the Interfaces Around Them
Inside Datapipelines: how pipelines are versioned as drafts and releases, how agent changes are tracked, and how REST, SSE, MCP, dashboards, and published endpoints reach the same pipeline.
A Semantic Layer for AI Agents: Inside Datapipelines
How Datapipelines gives AI agents a semantic layer with evidence-backed facts, inline metadata, reusable business definitions, and explicit review boundaries.
Pipelines, Part 1: The Contract Before the Code
Inside Datapipelines: how we specified pipelines before writing code. The pipeline contract, its type system, REST, SSE and MCP interfaces, version history, and the planning discipline behind them.