⏱️ Lectura: 10 min
Out of every 100 billion people born throughout human history, a website just picked one for you at random: with a birth year, a place on the planet, and even a story with documentary sources at the bottom.
📑 En este artículo
It’s called Any Human Ever and it turns a classic demographic question (how many humans have ever lived) into an interactive tool that any developer can dissect to understand weighted sampling and historical data visualization.
TL;DR
- Any Human Ever randomly picks a year, a place, and a life among more than 100 billion people who existed.
- The birth year draw uses a logarithmic scale because population grew exponentially.
- The location is chosen based on historical population density, shown as a map with brighter zones.
- Each generated life includes a narrative story with documentary sources cited at the end.
- The project lives at anyhumanever.com and lets you redraw year, place, and life separately.
- The 100 billion figure matches, in order of magnitude, classic demographic estimates like those from the Population Reference Bureau.
Introduction
Any Human Ever starts from a simple premise: humanity, across its entire history, added up to more than 100 billion births, according to the project itself at anyhumanever.com. The figure matches, in order of magnitude, classic demographic estimates like the ones published by the Population Reference Bureau for decades. Instead of showing that figure as a cold statistic, the site turns it into an experience: you click a button and the system draws, step by step, a hypothetical but statistically plausible life within that universe of 100 billion people.
What Happened
The site’s flow has three steps, each with its own draw. First, a birth year is chosen: almost everyone who ever lived was born recently, because world population grew exponentially and a random life is far more likely to fall near the present than in antiquity. Second, a geographic location is chosen on a map where brighter zones indicate higher historical population density. Third, with those two data points fixed, the system builds a life and displays its story, with a final section of documentary sources backing each narrative.
flowchart TD
A["Draw birth year (log scale)"] --> B["Confirm year"]
B --> C["Draw location by historical density"]
C --> D["Confirm location"]
D --> E["Generate life and read its story"]
Context and History
The question of how many people have ever lived is a demography classic. The usual way to answer it is to reconstruct, century by century, birth rates and estimated population sizes, and add up the accumulated births since Homo sapiens appeared. The result depends on the assumptions used for prehistory, where there are no censuses: that’s why different sources arrive at figures between 100 and 120 billion. Any Human Ever sits at the conservative end of that range and states so explicitly on its homepage.
What’s interesting for a developer isn’t so much the final number, but the pattern behind it: human population growth wasn’t linear. For most of history it grew very slowly; only in the last two or three centuries did it spike exponentially, a phenomenon documented on the Wikipedia page on world population. That exponential curve is, literally, the probability distribution Any Human Ever uses to draw the birth year.
💭 Key point: if population grows exponentially, the vast majority of the 100 billion births accumulated throughout history happened in the last few centuries. That’s why a uniform draw by year would produce an unrealistic result: you need to weight by population, not by number of years elapsed.
Technical Details and Performance
To represent that curve without ancient centuries disappearing from the chart, the site offers an explicit toggle between logarithmic and linear scale on the years-before-present axis. It’s a very common data visualization decision when a variable spans several orders of magnitude:
| Time axis scale | When it’s useful | Advantage | Limitation |
|---|---|---|---|
| Linear (real years) | Comparing short ranges, e.g. 1900-2026 | Easy for humans to read | Compresses thousands of years of ancient history into a few pixels |
| Logarithmic (years before present) | Showing all of human history on a single chart | Gives proportional visual space to each era, even with less population | Distorts the perception of duration: an ancient century takes up the same space as a recent decade |
The mechanism that makes it possible to draw a year weighted by population is weighted random sampling, a classic algorithm any developer can implement in a few lines. The idea: instead of drawing a uniform index, you draw a point within the cumulative sum of weights (population by year) and find which segment it falls into.
function pickBirthYear(populationHistogram) {
const totalBirths = populationHistogram.reduce(
(sum, point) => sum + point.estimatedPopulation, 0
);
let threshold = Math.random() * totalBirths;
for (const point of populationHistogram) {
threshold -= point.estimatedPopulation;
if (threshold <= 0) return point.year;
}
return populationHistogram[populationHistogram.length - 1].year;
}With a histogram like [{ year: -8000, estimatedPopulation: 5000000 }, { year: 1950, estimatedPopulation: 2500000000 }, { year: 2020, estimatedPopulation: 7800000000 }], this function returns years close to the present far more often than ancient years, because each segment weighs proportional to its population, not its duration in years.
The same principle works for location, only weighting by population density instead of by year:
import bisect
import random
def pick_location(density_by_region):
regions = list(density_by_region.keys())
weights = list(density_by_region.values())
cumulative = []
total = 0
for weight in weights:
total += weight
cumulative.append(total)
point = random.uniform(0, total)
index = bisect.bisect_left(cumulative, point)
return regions[index]With a dictionary like {'East Asia': 1500000000, 'Western Europe': 200000000, 'Sub-Saharan Africa': 1100000000}, pick_location returns ‘East Asia’ more often than ‘Western Europe’, just like Any Human Ever’s brightness map favors the most densely populated zones. To confirm that weighted sampling like this actually works, you don’t need to read the code twice: just run it 100,000 times, count how many times each option came up, and compare it against the expected relative weight. If a region weighs 55% of the total, it should show up in roughly 55% of the 100,000 runs.
How to Try It
Trying Any Human Ever doesn’t require installing anything: it’s a website that runs in the browser, at anyhumanever.com. The flow is: first draw a year (and redraw it as many times as needed until you’re satisfied); then confirm it and move on to the location, draw it and confirm it too; finally, generate the life and read its full story with cited sources.
For a developer who wants to reproduce the core mechanic, not the whole site but the weighted sampling pattern, the two code snippets from the previous section are enough: one for the time axis, weighted by estimated population per year, and another for the geographic axis, weighted by population density per region. Both algorithms run in linear time relative to the number of histogram segments, so even with thousands of segments the draw is instant in any browser.
Impact and Analysis
The value of Any Human Ever isn’t in technical complexity (weighted sampling is a textbook algorithm), but in the experience design: it turns an abstract statistic into a personal narrative. It’s the same principle used by one-dot-per-person visualizations that represent censuses or migrations: numbers stop being numbers when the user receives just one, chosen for them.
For the web development community, the project is also a case study in how to communicate sampling bias without statistical jargon. Explaining that almost everyone who ever lived was born recently is, at its core, an accessible way to describe a distribution skewed toward the present, without using that terminology.
The project’s honest limit is the same one any historical demographic reconstruction has: the further back in time, the less hard data there is and the more the result depends on models and assumptions. Any Human Ever doesn’t claim that each life is the record of a real person; it’s a statistically coherent simulation, not a genealogical database.
⚠️ Heads up: the lives Any Human Ever generates are statistical reconstructions, not records of real, identified people. The further back the drawn year is, the more the result depends on demographic models and the less on direct census data.
What’s Next
The site itself leaves the door open to keep drawing: each step has its own button to redraw the result, and at the end you can start over with a new life. As an open data and visualization project, its natural path forward is to add more historical sources by region (each finished story already lists its documentary sources) and refine the geographic resolution of the density map.
Try it yourself: go to anyhumanever.com and draw your own year, place, and life in under a minute, with nothing to install.
📖 Summary on Telegram: View summary
Frequently Asked Questions
What is Any Human Ever?
It’s an interactive website that draws, step by step, a hypothetical life among the more than 100 billion people estimated to have been born throughout human history, combining a birth year, a geographic location, and a story with sources.
Where does the figure of 100 billion people come from?
It’s a demographic estimate based on reconstructing historical birth rates century by century. The project uses it as a baseline, and it matches, in order of magnitude, figures published by organizations like the Population Reference Bureau.
Why is the birth year drawn on a logarithmic scale?
Because world population grew exponentially: the vast majority of historical births happened in the last few centuries. A logarithmic scale makes it possible to show all of human history on a single chart without recent centuries visually crushing the ancient ones.
How does it choose where that person lived?
With the same principle as the year: weighted sampling, in this case by historical population density. The site’s map lights up more brightly in the zones that held the most inhabitants in each era.
Are the generated stories real or made up?
They’re statistically plausible reconstructions, not records of identified people. Each story cites the documentary sources used to build it, but it doesn’t correspond to a verifiable historical individual.
Do I need to install anything to try it?
No. Any Human Ever runs entirely in the browser at anyhumanever.com, with no signup or installation required.
References
- Any Human Ever: the project’s original site, with the year, location, and life draw flow.
- Population Reference Bureau: a reference organization for estimates on how many people have lived throughout history.
- Wikipedia: World population: context on the exponential growth of world population.
- United Nations: World Population Prospects: reference source for historical demographic data and projections.
📱 Enjoying this content? Follow @programacion on Telegram for daily tech content in Spanish: quick summaries, fresh content every day. @programacion
Imagen destacada: Foto de Leandro Barreto en Unsplash
0 Comments