Research queue
Loading public-safe aggregate progress…
CT-04 // Phase 3 construction
Build the weights before drawing the leaderboard.
The chronology and mechanism ontology are already frozen. This phase converts that research into historically appropriate cultural nodes, dimension scores, O/C/S/A roles and eventually relative weights — without projecting modern countries backward or letting the graphics decide the answer.
Current state
No frozen entity weights yet.
Source-derived candidate queues now exist for the public 1860, 1950 and 2026 anchors, but candidates are kept private until selection and coverage audits are complete. The live Timeline therefore remains in clearly labelled UI PROXY mode for entity size/order while its mechanism cards use the public frozen data.
Live research progress
Evidence first, scores later.
The counters below are deliberately aggregate-only. Draft node identities and individual draft scores stay private until the research pass is reviewed and frozen.
Historically appropriate entities
Use states only when the state is actually the useful carrier. Nodes may instead be civilizational systems, religious networks, diasporas, interaction spheres, colonial systems or information networks.
Score the parts separately
Population, economy, trade, technology, military reach, information, law/admin, science/education, religion, language/arts/custom, contemporary influence and inherited transmission are scored separately before any aggregate exists.
Attribution survives aggregation
Origin, Carrier, Sponsor and Amplifier roles stay explicit. A later carrier or broadcaster does not retroactively become the origin of what it transmits.
Look where archives are weakest.
Every anchor carries a mandatory geographic/network coverage checklist so empire size, surviving texts and modern visibility cannot silently decide which systems enter the model.
N/A is not zero.
A dimension can be applicable, structurally not applicable, or blocked by insufficient evidence. Missing evidence and non-territorial carrier types are never silently converted into zero scores.
Evidence → rationale → score.
Every draft ordinal score requires an evidence note and a separate scoring rationale. Draft values remain private and do not drive the public Timeline until the full anchor is reviewed and frozen.
Working ordinal rubric
0–5 is not a percentage.
Dimension review uses an era-relative ordinal scale. The final relative weight is a later calculation and will be tested under multiple weighting profiles. A score of 5 means “first-tier for this dimension at this snapshot,” not “five times as important as a 1.” A zero is reserved for a defensible absence/unsupported contribution — it is not the fallback for missing evidence.
Score rubric
Loading the public scoring bands from Supabase…
Sensitivity phase
One formula does not get to decide history.
Once dimension scoring is frozen, the model will run alternative profiles that vary demographic weight, economic/material weight, contemporary reach and inherited transmission. A result becomes interesting when its broad shape survives those changes.
Hard stop rules
Things this phase refuses to do.
No backward credit from future descendants. No state-death = culture-death shortcut. No “largest empire therefore largest culture” rule. No filling weak-source regions with zeros merely because the archive is thin. No converting a smooth slider animation into fake annual evidence.