Hacker News Reader

As an ex-CITP member, that's most definitely a human written joke.

Maybe. But this article was at the least drafted by claude. It does all the things, including that weird word density + concept vapidity that makes parsing so difficult and annoying.

The report does feel like LLM output.

The biggest problem I see related to that is the marked haziness over what exactly has been accomplished. As best I can tell, this paper claims to have deanonymized primary election ballots that were cast during the early voting period. But it's written as if it was deanonymizing general election ballots cast on election day.

Human authors really need to screen their output with a (relatively) good AI detector.

It's surprising that Pangram gives 0% AI text. There are notable examples of LLM writing here:

> Tied to a specific ballot is the stronger claim and requires one additional public record.

clearly a claude-ism

I guess the author went back and forth running pangram and slightly tweaking the text until it read "0% LLM text"