Paper Taxonomy — two axes
A mental model for sorting TerraPulse papers (Mike, 2026-06-09). Most "how do you categorize the papers" cuts collapse into two independent axes. Plot a paper on both and you know how to treat it.
Axis 1 — Data velocity (how fast the underlying source changes)
This is inherited from the data layer, not the paper. You know it before you write a word.
- Fast-moving — live event streams. New rows land hourly or faster. Examples: USGS earthquakes, NWS alerts, CNEOS fireballs, GLM lightning.
- Slow-moving — periodic aggregates. New rows land monthly or yearly. Examples: ERA5 monthly means, USDM drought index, World Bank climate-risk, tree-cover / carbon.
What it tells you: the natural refresh cadence. A result built on fast data is worth re-running often (it can read different next week). A result on slow data is stable between refreshes (re-run it next week, same number).
Axis 2 — Window type (does the question ever finish?)
This is a choice the author makes, not a property of the data. It's set by where you draw the fence around the question.
- Closed window — the question is bounded to a fixed past interval. "Did fireballs surge in 2026?" Once the window elapses, future data falls outside the question and can't move it.
- Open question — no end date. "Is there a quake / UFO-report coincidence?" Every new event is one more trial. It's a work-in-progress by construction.
What it tells you: whether reruns ever stop. Closed → the paper can reach a definitive end. Open → it's perpetual.
The "finalized" caveat
A closed window isn't truly done until the source stops revising inside it. USGS revises magnitudes; ERA5 doesn't finalize a month until ~3 months later.
Definitive end = window closed AND revisions settled.
The scoping lever
Almost any perpetual question can be converted into a finishable paper by bounding it. Same data, same method:
- "Do quakes correlate with UFO reports, all-time" → forever-WIP.
- "...over 1995–2014" → closeable.
The fence is the lever. Reach for it when you want a paper to ship and stay shipped rather than become a standing monitor.
The 2×2
| Fast-moving data | Slow-moving data | |
|---|---|---|
| Closed window | Finishes; data was perishable but the fence froze it. Once finalized, lock it. (e.g. fireball-surge-2026) | Finishes and is stable the whole way. Shelve and trust. |
| Open question | Perpetual; high-value standing monitor, re-run often. (e.g. quake/UFO coincidence, all-time) | Perpetual but slow; re-run on the slow cadence. |
Practical payoff (how to treat a published paper)
- Velocity → shelf life. Fast-data snapshots are perishable: tag them "re-run me on cadence X." Slow-data snapshots are stable for a year+: forget about them.
- Window → does watching ever stop. Closed + finalized: archive it, done. Open: it's a standing instrument, not a finished artifact, treat it as infrastructure to maintain.
- Most of what we've shipped is closed-window snapshots (earthquake-lights null, Skinwalker cross-match, fireball surge). The open-question / standing-monitor quadrant is mostly empty, and the platform (60 s ingestion, forever) is built precisely for it. That's a gap worth filling deliberately, not by accident.