AI anxiety and AI joy, sometimes from the same person

"The last time my family was replaced by technology" drew a quiet, reflective thread. One commenter's grandad ran a letterpress shop and, when phototypesetting arrived, didn't retrain. He just took smaller jobs until it was wedding invitations. Another said their great-great-grandpa thought cars would end horse jobs, and he was right. In the ML PhD in Japan thread, the top comment says to worry more about AI automating your PhD than about where you do it. The 60-year-old who says Claude Code re-ignited a passion got replies from a 51-year-old electrical engineer going solo-founder and a person with bipolar disorder who calls it an accessibility tool. One dissenter said using Claude all week felt unfulfilling, like getting an A on a test you didn't take.

The math story sharpens the economics. A paper on responsible release of AI-generated mathematics asks AI companies to stop testing hard problems on proprietary models and to pay human mathematicians to understand AI output. Commenters called that gatekeeping, or an unusual ask. The key bit: the verification and interpretation work is where humans are being pushed, and nobody has agreed who pays for it.

The mood isn't doom. It's people noticing they have two choices, like the letterpress grandad: ride the new tool or shrink into a niche.