The Wayfinders' Atlas

A single-file, offline world model and voyaging simulation of how Austronesian/Polynesian peoples crossed the Pacific — from the homeland off Taiwan out to the three corners of the Polynesian Triangle (Hawaiʻi, Rapa Nui, Aotearoa).

▶ Open it: index.html — double-click it from a fresh clone. No build step, no server, no API keys, no network required.

Mirrored artifact (same file): public/artifacts/polynesian-spread/cursor-fable-5-thinking-high.html, served on the results site at /artifacts/polynesian-spread/cursor-fable-5-thinking-high.html.

What it is

The page is three layers stacked into one inspectable model:

1. A world model. 21 island nodes from the real record — each with a settlement-evidence window (radiocarbon / tephra / dated proxies) shown as a range, not a false-precision point, plus a sourced note. A "World model" tab narrates the spread in six movements; a "Debates" tab surfaces the genuinely contested pieces; a "Sources" tab lists the primary literature.
2. A physics layer. A deliberately coarse, legible Pacific oceanography (SE/NE trade winds, the equatorial counter-flow, ENSO westerly windows, and the major surface currents) combined with a canoe polar (speed, range, windward ability) that scores every inter-island crossing.
3. A solver. Given fixed voyaging assumptions, a shortest-arrival-time search (Dijkstra over the voyaging network) computes when — if ever — the spread reaches each island. The result is rendered as an animated map, a timeline against the evidence, a per-hop trace, a per-island inspector, and a verdict.

The headline question

What did the canoes and the voyaging strategy have to be capable of to reproduce the spread we actually observe, within the dates the evidence allows? The simulation answers it by being falsifiable. Four built-in presets:

• Weatherly two-way wayfinding — reproduces the record (all three corners, the long pause, the fast East Polynesian pulse).
• Drift voyaging only (Sharp 1956) — square sails, no windward ability, one-way: fails to enter East Polynesia at all, reproducing the classic result of Levison, Ward & Webb's 1973 computer simulation.
• Marginal canoes, small holds — reaches the East Polynesian core but strands the far corners (Hawaiʻi, Rapa Nui) for lack of provisioning range.
• Climate-driven (heavy ENSO windows) — modest boats, but frequent westerly windows do the upwind work: an alternative success path.

The takeaway you can defend after playing with it: pure downwind drift cannot explain the eastward push; reproducing the record requires either genuine windward ability with two-way (return-capable) voyaging or a climate regime that opens the eastward door often enough — and even then, enough provisioning range to span the final long jumps to the corners.

How to evaluate (matches the prompt's acceptance criteria)

1. Open the model. Double-click index.html. The map, winds, voyage tracks, and timeline render immediately (the success preset auto-runs and animates).
2. Read the world model. Use the right-hand tabs: World model (sequence + dates + citations), Debates (the long pause, EP tempo, upwind vs. drift, two-way voyaging & the Americas, Rapa Nui timing), and Sources.
3. Run a simulation. Drag the Windward ability, Provisioning range, Canoe speed, Navigation skill, Westerly windows (ENSO), or Voyaging cadence sliders, toggle two-way voyaging, or click a preset, then Run spread. Watch the verdict flip between success, "stalls / strands the corners," and outright failure.
4. Trace why. Turn on the wind, current, and failed-crossing overlays on the map; read the Run trace (every hop with its limiting physics — distance, point of sail, angle off the wind, landfall odds, window wait); and click any island to open the Inspect tab, which lists every incoming route and exactly why it is feasible or blocked.
5. Timeline vs. record. Grey bars are the evidence windows; diamonds are this run's arrival dates, coloured hit / too-early / too-late / never. The West Polynesian "long pause" band is drawn in.

Files

• index.html — the complete deliverable (world model + simulation; vanilla JS, HTML5 canvas, no dependencies).
• _tune.mjs — a throwaway Node calibration harness used to tune the physics constants against the milestone windows. Not needed to view or evaluate the page; run with node _tune.mjs to see the regime table on the command line.

Honesty about the model

The oceanography is an analytic, hand-built field, not a reanalysis dataset, and each crossing is reduced to an expected arrival time — so it is built to be legible and falsifiable, not numerically authoritative. The robust outputs are the qualitative regimes (drift fails; weatherly two-way voyaging with wind windows succeeds; marginal range strands the corners), which match the cited literature. Dates carry the uncertainty the evidence carries; see the Sources tab for the full list.