About
Who we are and how to join
Cohese is a small, independent research lab for computational political geography. We study how people vote, what turns out to be predictive of how they will vote in the future, and which methods make those questions tractable at the scale of every precinct in the country.
The work tends to start from something noticed in election data and end as a long write-up with the code behind it. Shoal samples neutral redistricting plans on a GPU fast enough to draw every map, everywhere, at once; DeepSky estimates precinct partisanship from satellite imagery and census data. Both were built on the Apple silicon Macs we already have, because Cohese runs on very little money, and the Inference page says more about how we find compute for the bigger ideas.
Join us
We are a few people, mostly friends and other young people who care about the same questions, and nearly every project here has passed through several of our hands. We would like the group to keep growing.
If you want to do something under Cohese, reach out. We would love to host your paper, help spread the word, or work on something together, and we have resources to share: a few Apple silicon machines on our tailnet, Mac minis among them, that we use but rarely to the limit. If you have a question you want to chase or research that needs those machines, get in touch.