Skip to content

Latest commit

 

History

History
32 lines (19 loc) · 1.78 KB

SHARING.md

File metadata and controls

32 lines (19 loc) · 1.78 KB
created 2026-02-18

DataPressr

Turn raw data finds into published datasets — fast, with AI.

The goal: go from "interesting data thing" to a polished, published entry on DataHub in under 10 minutes. No more half-finished tabs and good intentions.


Where We Are

We're at the end of Stage 1: the posting loop exists, on paper and in practice.

Done:

  • Plan — a clear, staged approach: backlog first, manual posts next, automation only after we feel the friction
  • Backlog — 100+ dataset candidates imported from the Data Commons issue tracker, plus a live inbox for new finds
  • Data model docs — defined the data hierarchy, lifecycle, and a catalog-as-repo pattern (with EH.net as a concrete example)
  • Dashboard candidates — an initial list of data stories worth building

Update: Stage 1 worked — 20+ datasets published to DataHub from the backlog. /init//validate//push cover the last mile, and structure (turning a raw source into a clean, typed dataset) is now a real skill too — see docs/plans/skills-vision.md. /validate and the structure playbook's cleanup idioms are backed by an actual test suite (npm test), not just prose.

The backlog/inbox/dashboards markdown files are superseded. Beads (bd ready) tracks actionable work; GitHub issues retain lightweight captures and historical discussion. Start a working session with NEXT.md.

Next: capture/archive skills, then a few data stories written by hand before deciding on a charting approach — see docs/plans/skills-vision.md.


The insight driving this: don't design the system before you've felt the workflow. Ten manual posts will teach more than ten hours of architecture.