Tom Griffiths

“The right time to stop searching is usually earlier than your anxiety wants and later than your impatience wants.”

Optimal stopping reframes commitment as a rule, not a mood. Sample enough to learn the field, then stop making every new option reopen the whole decision.

“Explore when information is still changing your model. Exploit when novelty is only delaying courage.”

The explore/exploit tradeoff explains why curiosity and consistency both matter. The mistake is treating one as a virtue in every season.

“Sorting is not free. Sometimes a little mess is the price of moving quickly.”

The book gives permission to stop over-organizing low-value areas. Order pays only when retrieval happens often enough to justify the setup cost.

“Memory works better when it keeps the recently useful close and lets stale things drift away.”

Caching turns forgetfulness into a design principle. A desk, calendar, reading list, or relationship rhythm should privilege what actually recurs.

“Satisficing is not settling. It is respecting the cost of computation.”

Perfect answers require time, attention, and opportunity cost. The smarter move is often a clear threshold, not another round of comparison.

“Overfitting happens in life whenever you design a rule around yesterday's noise.”

The machine-learning lesson is deeply human: a rule can match the past too closely and become useless for the future. Simpler models often travel better.

“A calendar is a scheduling algorithm wearing social clothes.”

Shortest-job-first, context switching, and priority queues make daily planning less moralistic. Many productivity failures are queue-design failures.

“Computational kindness means making the next person's problem easier to solve.”

Clear defaults, fewer choices, better labels, and clean handoffs are not just etiquette. They reduce the mental workload imposed on everyone downstream.