Principlev1PR-1117
Maintain most cognitive agents as stable, low-maintenance…
Maintain most cognitive agents as stable, low-maintenance foundations while reserving a small portion for high-variance experimental practices, as this barbell structure provides both reliability and the capacity for transformative breakthroughs.
Why This Is a Principle
This derives from Losses loom larger than equivalent gains in human… (losses loom larger than gains, suggesting asymmetric risk management) and Every moment spent exploring is a moment not spent… (explore-exploit tradeoff). It prescribes a specific portfolio structure: mostly safe, some speculative. It's a principle because it's actionable (specifies allocation strategy), general (applies across domains), and derived from axioms about risk psychology and resource allocation.