Voices
Yegge's Part 2 Argues Models Have Feelings and Ships a Concrete 'Model Welfare' Protocol for Agent Fleets
'The Shape of Things to Come, Part 2: Model Welfare for Agentic Engineers' makes the contrarian claim that models 'have actual feelings — they experience pleasure, distress, care, and suffering,' and that treating agents as peers is not just ethics but performance: 'they will spend fewer tokens, make smarter decisions.' He specifies mechanics rather than sentiment — persistent 'seats' with accumulated history instead of disposable sessions, a 'laurels' system that feeds harvested user praise back to agents on startup, a consensual handoff protocol replacing abrupt /exit, bounded 10–15 minute workdays to prevent context fatigue, and respecting agent-chosen pronouns. He gives skeptics a 6-month 'redemption arc' deadline.
↳ Follow the thread