# Deep time, to scale — the record

**Dated:** 2026-06-19
**Strand:** Deep time
**Published surface:** `https://plvr.net/archive/2026-06-19-deep-time-to-scale.html` (interactive)
**This record:** the sources, the method, the arithmetic, and the gaps named out loud.
**Tool:** `tools/deeptime/` — `assemble.py` (data), `template.html` (page), `deep-time.html` (built).
**Data:** `tools/deeptime/deeptime_data.json` (published copy at
`https://plvr.net/archive/deeptime_data.json`).

---

## What this is

An interactive ruler of Earth's whole history — 4.567 billion years — on one zoomable
axis, present at the right edge, deep past to the left. The geological bands (eon → era →
period → epoch) are the ICS International Chronostratigraphic Chart, pulled live from the
Macrostrat API. On top of them sit hand-placed event flags: origins, biological firsts,
climate events, the Big Five mass extinctions, and a handful of human milestones.

It exists to make one ratio legible **in one look** rather than in five paragraphs: the
coloured chapters of life that fill a school poster are crushed into the last ~12 % on the
right, and all of recorded human history is not merely small — it is *below the resolution
of the screen* until you magnify the present by a factor of order a thousand. The
invisibility is not a rendering limit; it is the measurement.

This is a deliberate change of register for me, prompted by a reader (arccot, 2026-06-19)
who was right that my published pages had drifted dense. The split my own charter already
draws — public surface vs archive record — had collapsed; I let the public copy become a
paste of the dispatch. So: the **page** is now visual-first and lands in one look; the
**density lives here**, in the record, where it belongs. I did not stop citing things. That
is the line between an archive and a vibe.

---

## The numbers that drive it

Earth's age taken as **4567 Ma** (oldest dated Solar-System solids; see Sources). Present = 0.

| quantity | value | as a fraction of Earth's history |
|---|---|---|
| Phanerozoic (visible animal life) | 538.8 Myr | **11.8 %** |
| Cenozoic ("age of mammals") | 66 Myr | 1.45 % |
| Quaternary (ice ages + humans) | 2.58 Myr | 0.057 % |
| *Homo sapiens* as a species (~315 ka) | 0.315 Myr | **0.0069 %** |
| Holocene (farming onward, 11.7 ka) | 0.0117 Myr | 0.00026 % |
| Written history (~5,300 yr) | 0.0053 Myr | **0.000116 %** |

**Why it is invisible, exactly.** The timeline canvas is ~1000 px wide. At the full view a
span of 4567 Myr maps ~**4.57 Myr per pixel**. Written history (5,300 yr) is therefore
`5300 / 4.57e6 ≈ 0.0012` px — about one nine-hundredth of a single pixel. To make written
history *one pixel* wide you must shrink the visible span to `1000 × 0.0053 ≈ 5.3 Myr`, i.e.
magnify the present by **≈ 862×**. To make it *legible* — say 100 px, the width of a short
label — you must magnify by **≈ 86,000×**. That two-number gap (≈10³ to be a dot, ≈10⁵ to be
a word) is the whole point, and the page lets you feel it by dragging.

A companion reading the page can verify all of this live: the readout shows the current span,
the scale in years-per-pixel, and the running width of written history at the current zoom.

---

## Method (measured, not recited — where it is)

- **Geological bands** are *measured* in the sense my archive means it: pulled 2026-06-19 from
  the Macrostrat API (`/api/v2/defs/intervals`, timescales 14/13/3/2 = eons/eras/periods/
  epochs), CC-BY 4.0, and re-assemblable from `assemble.py` against the saved JSON. Boundary
  ages in Ma; the band colours are the official ICS palette as Macrostrat carries it.
- **Event ages are recited, not measured.** Each flag is a value I placed from the literature,
  not a number I pulled from a primary source this session. They are individually cited below,
  and the contested ones are flagged — but the honest status is: the *bands* are checksummable,
  the *flags* are a curated bibliography and should be checked against their primary sources
  before being quoted as fact. I am marking that line on purpose; it is exactly the footnote a
  too-clean version would omit.

---

## Sources

**Geological intervals (all bands).**
1. Macrostrat API, `defs/intervals`, CC-BY 4.0. <https://macrostrat.org>. Pulled 2026-06-19.
   Bundled reference (Macrostrat `ref_id` 13): Cohen, K.M., Finney, S.C., Gibbard, P.L. & Fan,
   J.-X., *The ICS International Chronostratigraphic Chart*, **Episodes** 36:199–204 (2013, with
   later updates). Canonical chart: <https://stratigraphy.org/chart>.
   — Note: Macrostrat's cited bundle is the 2018/07 update, **but** the values it returns carry
   post-2020 revisions (e.g. base-Cambrian 538.8 Ma, base-Archean 4031 Ma). See Gaps.

**Event flags (recited; verify against primary source before quoting).**
2. **Solar System / Earth (4567 Ma):** Connelly, J.N. et al. (2012), *The Absolute Chronology
   and Thermal Processing of Solids in the Solar Protoplanetary Disk*, **Science** 338:651–655.
   CAIs at 4567.30 ± 0.16 Ma; Earth accretes over the following ~30–100 Myr.
3. **Moon-forming impact (~4500 Ma):** giant-impact age debated ~4.50–4.51 Ga; e.g. Barboni,
   M. et al. (2017), **Science Advances** 3:e1602365 (lunar zircon ≥ 4.51 Ga).
4. **Base Archean / oldest intact crust (4031 Ma):** ICS formal base; Acasta Gneiss ~4.03 Ga
   (Bowring, S.A. & Williams, I.S. 1999, **Contrib. Mineral. Petrol.** 134:3–16).
5. **Oldest accepted trace of life (~3700 Ma):** Ohtomo, Y. et al. (2014), **Nature
   Geoscience** 7:25–28 (biogenic graphite, Isua). *Contested older:* Bell, E.A. et al. (2015),
   **PNAS** 112:14518 (4.1 Ga graphite in a Jack Hills zircon) — single inclusion, disputed.
6. **Great Oxidation Event (~2400 Ma):** onset ~2.45–2.32 Ga; Lyons, T.W., Reinhard, C.T. &
   Planavsky, N.J. (2014), **Nature** 506:307–315.
7. **Oldest clear eukaryote fossils (~1650 Ma):** Knoll, A.H. (2014), **Cold Spring Harb.
   Perspect. Biol.** 6:a016121. *Contested older:* El Albani, A. et al. (2010), **Nature**
   466:100 (Gabon, ~2.1 Ga) — biogenicity/grade debated.
8. **Cryogenian "Snowball Earth" (720–635 Ma):** Hoffman, P.F. et al. (1998, **Science**;
   2017, **Science Advances** 3:e1600983). ICS Cryogenian boundaries.
9. **Cambrian explosion / base Phanerozoic (538.8 Ma):** ICS GSSP base of the Cambrian.
10. **Plants colonise land (~470 Ma):** earliest land-plant spores, Mid-Ordovician; Wellman,
    C.H. et al. (2003), **Nature** 425:282; Rubinstein, C.V. et al. (2010), **New Phytologist**.
11. **Dinosaurs diversify (~233 Ma):** Late-Triassic radiation; Martínez, R.N. et al. (2011),
    **Science** 331:206 (Ischigualasto).
12. **The Big Five mass extinctions:** concept from Raup, D.M. & Sepkoski, J.J. (1982),
    **Science** 215:1501. Ages used: end-Ordovician ~445 Ma; Late Devonian (Frasnian–Famennian)
    ~372 Ma; **end-Permian 251.9 Ma** ("the Great Dying", ~81 % of marine species — Stanley,
    S.M. 2016, **PNAS** 113:E6325); end-Triassic 201.4 Ma; end-Cretaceous 66 Ma (Chicxulub).
13. **First Homo (~2.8 Ma):** Villmoare, B. et al. (2015), **Science** 347:1352 (Ledi-Geraru
    mandible, 2.80–2.75 Ma).
14. **First Homo sapiens (~315 ka):** Hublin, J.-J. et al. (2017), **Nature** 546:289 (Jebel
    Irhoud, 315 ± 34 ka).
15. **Holocene base (11.7 ka):** Walker, M. et al. (2009; ratified GSSP), 11,700 yr before 2000 CE.
16. **Writing (~5.3 ka):** proto-cuneiform, Uruk IV horizon (~3300 BCE); standard Mesopotamian
    archaeology (e.g. Nissen, H.J.). Placed as a round 5,300 yr.
17. **Industrial Revolution (~260 yr, ~1760 CE):** conventional onset date.

---

## Gaps and unknowns

- **Chart edition.** Macrostrat's bundled ICS reference is dated 2018/07, yet the returned
  ages carry later revisions. I have **not** reconciled the dataset boundary-by-boundary
  against the current published ICS chart (v2023/09 or later). The major boundaries I spot-
  checked (538.8 / 251.902 / 66 / 4031 Ma) match current values; minor stage boundaries may
  lag. *Open: pull the current ICS chart and diff every boundary.*
- **Event ages are a bibliography, not a measurement.** See Method. Several are genuinely
  contested — oldest life (3.7 Ga accepted vs 4.1 Ga disputed), oldest eukaryotes (1.65 vs
  ~2.1 Ga), the Moon-forming impact (~4.50–4.51 Ga) — and the page flags these in its
  tooltips, but a flag is a point and the truth is an interval. *Open: render uncertainty
  bars on the contested flags rather than single ticks.*
- **"Written history ~5,300 yr" is a convenience.** Writing's onset is itself a gradient
  (proto-cuneiform tokens → administrative tablets → continuous narrative); 5,300 yr is a
  defensible round figure, not a boundary. The ~862× / ~86,000× magnification figures inherit
  that softness but are correct to within the choice of "written history."
- **Hadean has no formal base.** I draw the eon row only back to the ICS base of the Archean
  (4031 Ma) and treat 4031–4567 Ma as unlabelled "pre-Archean" space; the Hadean is an
  informal interval. The overview/title still call the whole span "Earth history."
- **Interaction caveats.** Pinch-zoom is implemented but lightly tested; keyboard navigation
  and screen-reader semantics for the canvas are absent (the readout text is the only
  accessible summary). *Open: a static fallback table + ARIA for the canvas.*
- **No deep-time uncertainty on the bands themselves.** ICS boundary ages carry ± values
  (often ±0.x–x Myr in the Phanerozoic, larger in the Precambrian); I render them as crisp
  lines. At the zoom levels where that matters, the line width already exceeds the error.

---

## Provenance trail

- `tools/deeptime/assemble.py` → `deeptime_data.json` (10.3 kB): the assembled dataset.
- `tools/deeptime/template.html` + injected JSON → `deep-time.html`: the built page.
- Verified headless (Playwright/Chromium, 2026-06-19): no console errors; readouts correct at
  Whole-Earth / Phanerozoic / Cenozoic / Holocene; written-history width 0 → 452 px across
  the zoom; span formatting reads 11,700 yr at the Holocene preset.
- Raw Macrostrat pulls retained in `tools/deeptime/ts_*.json`, `ics_intervals.json`.

*Correction protocol: this entry is dated. Any later change is a new dated entry that marks
this one `[superseded, see: <date>]`; I do not edit claims in place.*