a test nobody would ever run
What does it cost to turn off the internet?
When a government cuts the internet, you can't compare that country to itself with the internet on. So I built each country a twin out of countries that stayed online, using GitHub's public data on how much code every country pushes each quarter.
01Build the twin
The twin is a weighted mix of countries that moved like this one before the shutdown. The gap after is the effect. Try the simple comparisons too, they can point the wrong way.
02All nine shutdowns
Long shutdowns show up clearly. Short ones get lost in the noise, part 05 shows why. The diamond is where the country would land if every day offline cost exactly one day of work.
03Does a day offline cost a day?
04Where did the activity go?
GitHub places each account where its connections come from. The number of developers placed in a country almost never falls, it just keeps growing. In the two longest shutdowns it fell, which is what you'd see if people switched to VPNs or left. So some of the drop is people still coding, just showing up somewhere else on the map.
Developers placed in the country vs the twin, in the shutdown quarter.
05Can quarterly data even see it?
I planted fake shutdowns of known size in countries that never had one, 250 times per size, and checked how often each method caught them. Slide to pick a length.
06What this can't tell you
Shutdowns come with coups, protests and war, so these numbers are the cost of everything happening at once. Part 03 is my way around that, it checks whether the days offline explain the drop on their own. Quarterly data also hides how fast things bounced back within a quarter. And the twin assumes the country would have kept moving like its donors, which matters most for Iran in early 2026, when most of the world jumped about 30%.