Doxxing appears to be an LLM superpower.

That said, surely my style when prompting differs a bit from what/how I write on social media but I'm using nostr:nprofile1qqsdy27dk8f9qk7qvrm94pkdtus9xtk970jpcp4w48k6cw0khfm06mss64u96 so if they hold what they promise, my prompts should all be in a flood of other prompts by others and my agents. Unfortunately I can't always sanitize my input to remove all PII like my project names or stuff but it's pretty decent.