AI is an existential risk for humankind. We need to stop its further development. I endorse a Treaty Preventing Artificial Superintelligence, and encourage readers to sign the Statement on Superintelligence by the Future of Life Institute and the Open Statement by Control AI.

That said, AI is also a powerful tool, and this spare-time project — containing thousands of maps at just its first iteration — would never have been possible without it. Most of the content here, including the code, is generated by Claude Code.

Counterfactual legislatures

What would this legislature look like if the rules for electing it were different?

Three countries so far, and two different questions. In the United States the constituencies are drawn by hand, so the question is what a map drawn without any political input would have produced. In Brazil and Spain the constituencies are states and provinces that nobody redraws, so the whole question is the arithmetic that hands them seats — and both countries put a floor under it that costs something measurable.

Every number on these pages comes out of a published input and code in the repository, and every section had to reproduce something already known before it was allowed to produce anything new. Those checks are named as they come up; they are the reason to believe any of the rest.

United States 65% of the measured pro-Republican bias removed by a map with no politics in it — and 77% of the Black-majority districts removed with it. Brazil 7.97× heaviest vote against lightest, under a constitutional floor that turns out to cost the party system almost nothing. Spain 4.13× heaviest vote against lightest, widening at every election since 2008 without a word of the law changing.

What this is for

Not a proposal. Splitline districts ignore everything people reasonably care about, and multi-member proportional representation is almost certainly a better answer. The point is that a politically blind map is a measuring stick: the difference between it and the enacted map is a minimum of how much the real map was shaped by something other than geography.

That question got sharper in April 2026, when the Supreme Court held in Louisiana v. Callais that a state may draw districts for any political purpose regardless of racial effect. The defence is now “our intent was partisan, not racial” — and a map drawn with no political input at all is the natural thing to measure it against.

Across the three countries one pattern keeps recurring, and it was not the expected one. The rules that are easiest to get indignant about are not the ones doing the damage. Brazil's eight-seat floor is a 16-fold violation of one-person-one-vote and costs the party system 13% of its disproportionality; pooling the country into one district costs it 84%. Spain's apportionment method could be swapped for a fairer one and nothing would move; its two-seat floor is worth 18 seats. And an American map drawn with no politics in it recovers two thirds of the partisan bias while destroying three quarters of the Black-majority districts. In each case the big, visible, arguable rule turns out to matter less than a quieter structural one next to it.

The longer aim is to carry this to more countries and electoral systems, and then ask whether legislatures elected under different rules would have voted differently on policies whose effects we can actually measure. The roll-call work on the United States and Brazil pages is the first step of that, and it is the part most likely to be wrong in ways the checks cannot catch.