Tenacity is a core bottleneck agents lift

Agents never tire or demoralize. They keep trying where a person would quit. This lifted stamina changes what is possible: an agent will work at a problem for 30 minutes and come out victorious. The leverage play is to give an agent success criteria and let it loop, rather than instructions.

Watching an agent work for 30 minutes on a problem and emerge victorious is a genuine ‘feel the AGI’ moment. But tenacity without direction wastes tokens. The practical pattern is to shift from imperative instructions (‘do X, then Y’) to declarative success criteria (‘write tests that pass, then make them pass’). The agent loops longer, explores more paths, and surfaces solutions a human would have abandoned from fatigue. Pair this tenacity with human review at the checkpoint: the agent tries everything, the human decides what to keep.