Spyglass is an interactive historical atlas that charts human civilization from the dawn of agriculture to 2020. Explore what was happening everywhere else — at any place, in any era — through maps, data panels, and an AI-powered guide.
Explore the Atlas →Spyglass is not an encyclopedia. It's an instrument — a tool for exploring macro-history across place and time. Every civilization in the dataset is scored across governance, economics, society, culture, science, and military dimensions, using detailed rubrics that make cross-civilizational comparison meaningful rather than arbitrary.
The core question Spyglass answers is deceptively simple: what else was happening? When you're reading about the Roman Republic in 200 BCE, what was going on in Han Dynasty China, Maurya India, and Mesoamerica? Traditional sources bury this context across hundreds of separate articles. Spyglass puts it all on one map, one timeline, one screen.
Coverage spans from roughly 10,000 BCE — when agriculture first enabled permanent settlements — through 2020, across twelve world regions from Western Europe to Oceania. Particular care has been taken to ensure robust representation of Sub-Saharan Africa, Southeast Asia, and other regions that are historically underserved in English-language references.
The way we consume information has fundamentally changed. We scan, swipe, and skim. Attention spans have compressed. And yet when it comes to history — one of humanity's richest subjects — the dominant experience is still static paragraphs on a wiki page or a 400-page textbook.
People have adapted by settling for shallow takes: a two-minute YouTube summary, a paragraph skimmed mid-scroll. The depth is gone, but nothing visual or interactive has replaced it. You're either wading through dense text or getting a surface-level gloss.
Spyglass bridges that gap. You don't have to read thirty Wikipedia articles to understand how the Abbasid Caliphate compared to Tang Dynasty China. You can see them side by side — their populations, their governance structures, their technological levels, their trade networks — all at a glance, and then dive as deep as you want through narrative arcs and scored dimensions. It's the richness of a reference work with the immediacy of a modern product.
Spyglass delivers information through three complementary channels that work in concert. Each stream reinforces the others — the map shows where, the panels show what, and the chat explains why. You can enter from any direction: click a city on the map, browse entities in the panel, or ask the chat to take you somewhere. All three stay in sync.
An interactive world map with cities sized by population and cultural zones shown as dashed regions. Scrub the timeline from 10,000 BCE to 2020 and watch settlements emerge, grow, and sometimes vanish. Click any city to open its full history. Zoom in and labels appear; zoom out for the global picture.
A four-panel system that unpacks every place and civilization in the dataset. Panel A shows city identity and metadata. Panel B renders population sparklines with entity control overlays. Panel C is a horizontal entity browser — scroll through every civilization that controlled a city across time. Panel D dives deep into scored dimensions, narrative arcs, and legacy lists across seven thematic tabs: governance, economy, society, culture, science, military, and legacy.
This is not a simple Q&A bot — it's a navigator that controls the atlas. Ask "Show me the Song Dynasty at its peak" and it jumps the timeline to 1100 CE, selects Kaifeng, and opens the entity panel. Ask "Compare the Aztec Empire to the Inca Empire" and it pulls both entity profiles with accurate scores. The chat reads from the live dataset — it doesn't guess, it queries. Every entity or place it mentions becomes a clickable link that drives the map and panels.
A vertical timeline with playback controls and speed settings. Watch twelve millennia of civilization unfold, or jump to any year.
Every city has a population history chart showing growth, decline, and which entity controlled it at every point in time.
Scroll through every civilization that controlled a city — from its earliest known settlement to the modern era — in a horizontal timeline strip.
Every entity is scored 1–5 on 12+ dimensions using published rubrics. See at a glance how the Aztec Empire's bureaucracy compares to Song China's.
Every entity-period has prose narratives across 7 themes — population, governance, economy, society, culture, science, military — grounding the numbers in human stories.
The AI assistant doesn't just answer — it drives the atlas. It can jump to years, select places, open entity panels, and retrieve live data to answer your questions with precision.
Spyglass marries place-based geography with entity-based political history, then enriches both with layers of contextual data. The result is a dataset that can answer both "What was this city like in 1200?" and "How did this empire compare to its contemporaries?"
Over 700 places — cities and cultural zones — with coordinates, founding dates, modern equivalents, and historical name changes. These are the fixed points on the map: cities that persisted for millennia and regions that shaped cultures.
A junction layer connecting over 1,000 unique civilizations to the places they controlled, and when. This powers the "Ruled by" timeline in every city panel — tracing Constantinople from Roman capital to Byzantine seat to Ottoman jewel.
The knowledge layer: 2,000+ entity-period records with 64 columns each. Quantitative scores, categorical classifications, narrative arcs, population estimates, trade connections, key sources, and legacy achievements. Every row is a snapshot of a civilization in a specific era.
All data is generated through a multi-pass pipeline: quantitative scoring against published rubrics, narrative enrichment grounded in those scores, and cross-entity validation to catch inconsistencies. Population timelines carry confidence ratings — "low" for archaeological estimates, "high" for census-backed figures — so you always know how much to trust a number.
Every 1–5 score in the dataset is assigned using a published rubric with labeled tiers, definitions, diagnostic questions, and historical examples. Scores are absolute — a "3" means the same thing in 500 BCE and 1900 CE. Era-relative context (how exceptional a score is for its time) is computed at display time so users can see both the absolute level and the percentile ranking.
How formal and structured is the state's administrative machinery?
How codified and consistently applied is the legal system?
How secure and continuous is the political order?
How significant is this entity in regional and global trade networks?
How advanced are financial instruments and institutions?
How possible is it for individuals to change economic class?
What degree of legal and social agency do women hold?
How rigid is the social hierarchy? (Note: higher = more rigid)
How widely available is formal education?
How prolific and influential is the civilization's cultural production?
What is the overall level of applied technology?
How advanced is the practice of medicine?
What level of mathematical knowledge and application?
How professionalized and capable is the military?
How far does this entity's diplomatic network extend?
How severe are external military threats? (Higher = more danger)
Spyglass uses AI to research, score, and generate narrative content for its historical dataset. The process is structured — every score is assigned against a published rubric, narratives are grounded in quantitative data, and multi-pass validation catches inconsistencies — but AI can make mistakes.
Historical population figures, governance scores, and narrative arcs should be treated as informed estimates, not definitive scholarly claims. Population data before 1500 CE carries particular uncertainty — our confidence ratings (low / medium / high) indicate how much weight to give each figure. Scores represent structured approximations anchored to rubric definitions, not precise measurements. We welcome corrections and actively maintain the dataset — if something looks wrong, let us know.