Mental Models for Decisions

The best alternative you give up when you choose something else.

Downstream consequences beyond the immediate, first effect.

Solve by asking: "How could this fail?" then avoid those failure modes.

Areas you understand well enough to make reliable judgments; expand deliberately.

Probability-weighted average outcome; compare options by EV (and risk).

Start with the typical frequency for similar cases before using case-specific details.

Update beliefs by weighting new evidence against prior probability.

Decide by the next unit: marginal cost vs marginal benefit.

Small gains repeated over time grow faster than linear improvement.

A small input that produces a disproportionately large output (tools, code, capital, people).

A cycle where outputs influence future inputs (reinforcing or balancing).

The single most limiting constraint that caps system throughput.

People respond to what is rewarded, not what is said.

Misaligned incentives between decision-maker (agent) and stakeholder (principal).

Improving one dimension often worsens another; name what you are optimizing for.

Reversible decisions can be made fast; irreversible decisions require higher certainty.

Keeping paths open has value when the future is uncertain.

Outcomes are not equal; focus on choices with limited downside and meaningful upside.

Gets stronger from stress/volatility (not just resilient).

Models are simplifications; do not confuse them with reality.

Look for interactions and dynamics, not isolated parts.

A minority of causes often drive a majority of results; find the leverage points.

Shared resources get overused when individuals optimize privately.

Assume a future failure and list plausible causes to prevent them now.

Record context, assumptions, and expected outcome; review later to calibrate judgment.