Waiting for a map click.
The population data comes from
GHSL - Global Human Settlement Layer, specifically, the GHS population grid (R2023) snapshot of 2020 at 1km resolution using Mollweide coordinates. I've filtered out all square kilometres with less than 1000 people so you won't find much outside of cities and towns. The metadata of the cities comes from the
GHS-UCDB R2024A - GHS Urban Centre Database 2025, specifically the GHS_UCDB_GLOBE_R2024A dataset. This means that city boundaries look different than what you are probably accustomed to, sometimes bigger and sometimes smaller than locals would expect. More metadata comes from
GeoNames, specifically the cities15000.zip file, and from the basic list of
SimpleMaps
Don't take this too seriously. The population numbers are estimates. They do not come from censuses.
Highlighting the highest counts is also a way to highlight artifacts in a dataset. I've censored some
squares from appearing in the "Top in ..." lists when the estimates look clearly wrong but there are still lots of egregious-looking numbers. IMO it looks like the estimates are especially wrong in deserts and moountainous regions.
Over 98% of the code was written by GPT-5.1-codex-max, GPT-5.2-codex and GPT-5.3-codex