logoalt Hacker News

soltanovtoday at 7:54 AM0 repliesview on HN

This is not only a fairness problem. It is an agent-memory problem: a system can mistake its own early choices for evidence.