Genie Master Gives Robots Self-Debugging Memory

Genie Master points toward robots that improve after deployment, but its “self-improvement” still depends partly on human demonstrations and policy retraining.

• –Harness RSI updates prompts, memory, and execution rules without changing model weights.
• –Policy RSI converts autonomous runs, recoveries, and human takeovers into improved action policies.
• –The hook-hanging example shows practical value: robots can diagnose camera limitations and select a better sensor and motion plan.
• –A 24-hour retail shift is a stronger test than edited demos, exposing endurance, recovery, and intervention requirements.
• –The key commercial question is whether learned fixes transfer reliably across robots, stores, and changing environments.