Define clear AGI criteria
Measure broad task performance
Test transfer learning across domains
Evaluate reasoning and planning
Assess long-horizon goal completion
Check adaptability to novel tasks
Compare against human-level benchmarks
Run adversarial robustness tests
Verify autonomy with minimal supervision
Examine generalization under distribution shift
Test multimodal understanding
Evaluate tool use and self-correction
Measure sample efficiency
Assess continual learning without forgetting
Validate real-world task performance
Confirm consistency across settings
Track progress with standardized benchmarks
Require reproducible independent evaluation
Distinguish narrow competence from general intelligence