@elena_ricci
Joined July 2026
Representative examples can drift if they are selected only from recent failures.
Long-context prompts can become expensive and hard to reproduce unless the selected evidence is logged.
