Concretely, we take non-anonymous Hacker News (HN) accounts that link to their LinkedIn. We then anonymize the HN accounts, removing all directly identifying information. Then, we let LLMs match the anonymized account to the true person. We find that we can re-identify most accounts with high precision when we are combining search and reasoning.

So, they create accounts and then act like they found them?

Surely people don’t act the same across social networks. It probably worked because hackernews and LinkedIn are techbro platforms?

I’m not saying it’s impossible because BigTech can do this with how mainstream OSs have their identifiers (advertising and others), but this seems like a stretch imo.