This right here. Try running prompt-engineering/injection automation with iterative adjustments and watch how easy it is to select tokens that eventually produc

This right here. Try running prompt-engineering/injection automation with iterative adjustments and watch how easy it is to select tokens that eventually produce the desired output, good or bad. It isn't AGI, its next token prediction working as intended.