Jon Green
banner
jongreen.bsky.social
Jon Green
@jongreen.bsky.social

Assistant professor, Department of Political Science, Duke University
jgreen4919.github.io

Political science 25%
Psychology 15%

every passage I’ve seen from this article is like finding a soldier who’s been hiding on an island for five years and doesn’t know that Woke War 1 is over and they lost
hilarious and insane article. full of the sorts of people you meet all the time in left activism, who combine neuroticism and narcissism to bring down any institution they're allowed into.
A Western Mass. lesbian bar required KN-95 masks. When it changed the rules, all hell broke loose. - The Boston Globe
Last Ditch Bar in Greenfield wanted to build community. They got a firestorm instead.
www.bostonglobe.com

it’s very small n so hard to say for sure but in theory the longer you’re out of power the better the fundamentals but also the stronger the electability pressures (real or imagined)
hilarious and insane article. full of the sorts of people you meet all the time in left activism, who combine neuroticism and narcissism to bring down any institution they're allowed into.
A Western Mass. lesbian bar required KN-95 masks. When it changed the rules, all hell broke loose. - The Boston Globe
Last Ditch Bar in Greenfield wanted to build community. They got a firestorm instead.
www.bostonglobe.com

it happened to me (toddler started picking the music in the car)

for my use case I needed to add a couple of frills (take the weighted average of response probs instead of the single likeliest response) but it matched agreement with human coders that I’d been using a lightweight gemini model for previously

I've got an ideological scaling task. The best (but expensive) way to do it was with pairwise comparisons. A cheaper alternative is to use the probability distribution from an open-weight model, but the results from Qwen weren't good enough to switch. Jev is, and took a ~$100 job down to $0.50.

jev: very useful!