Core systems

The load-bearing pieces of the stack — what the rest of the lab runs on.

Run-time

What stakeholders and readers see — data stories, dashboards, tours, and operator surfaces on the viewport.

Dashboards & instruments

Not every useful artifact is a public data story. Some are tools for seeing, steering, or physically correcting the craft itself.

Build-time

Data conduits that make the viewport trustworthy — including context layers, agents, anti-hallucination resources, and community tooling.

Semantic layer & meta context

Open schemas and landscape work that teach agents to interpret metrics, not just query them.

Warehouse trust

Empirical grounding — query-log mining, join patterns, and knowledge graphs built from real warehouse work.

Domain agents

Builder, analytics manager, and ensemble workflows — including deep reads on how agent memory actually behaves.

Shareable workflows

Token-spread tricks and pipeline commands — some published, some still workspace-only.

Context Engineering for Analytics Engineers

article ↗

The Substack piece that started the conversation — what happens when your semantic layer needs to explain itself to an LLM.

Adoption

Self-executing walkthroughs and shareable skill pages for onboarding others to the same toolchain.