- The future is hard to predict, engineer serendipity and anti-fragility.
- Effort is not proportional to results.
- Compounding is powerful, literally exponential, and consistently under-estimated.
- Binding constraints and slow variables (tolerance, biology, cognition, regulation, culture, expertise pipelines, infrastructure, coordination, fixed supply, physics, …) limit improvements, often acting as invariants.
- Moats are expensive-to-try + vague-to-verify. Cheap trying + clear verifying invites automation. Cutting either cost (simulators, instruments) is a strategic act - deliberate or hostile.
- etc.
Sources:
- Slow variables - Tolerance as the real constraint: https://claude.ai/chat/e7d1e20f-b7b7-4e1a-9cd8-d6b40be16079