Summary of "Algorithms to Live By"

Core Idea

Computational thinking solves recurring human problems: apply algorithmic patterns to dating, career, productivity, and relationships
Know when to stop optimizing: perfect solutions rarely arrive; "good enough fast" beats perfect-but-never

37% Rule: explore 37% of your options without committing, then accept the first one better than everything you've seen
Combat natural impatience—you'll stop too early otherwise
Apply to: apartment hunting, job interviews, dating

Young? Explore new careers, hobbies, people; unknown options deserve credit
Older? Exploit proven favorites; your time horizon shrinks, so experimentation costs more
Shift strategy gradually as you age, not abruptly

Copernican Principle: expect things to last as long as they already have (Berlin Wall lasted 8 years, expect 8 more)
Good predictions require good priors; without relevant experience, predictions fail
With small data, use your existing knowledge, not just raw numbers

Match metric to goal: Shortest Processing Time clears your to-do list fastest; Earliest Due Date minimizes missed deadlines
Weighted priority: divide task importance by duration; tackle high importance-per-unit-time first
Batch similar tasks (email blocks, bill-paying day) to minimize context switching

LRU (Least Recently Used): keep frequently accessed items physically close to where you use them
Multi-level organization works: closet, basement, storage unit by access frequency
Geography matters—store exercise gear by the front door, vacuum near the stairs

Sort only if you'll search repeatedly; unsorted is optimal for one-time lookups
Bucket Sort: rough-sort by major categories first, fine-sort later (saves effort)

When optimization is intractable, solve an easier version first (remove constraints, allow approximations)
A 90% solution in 1% of the time beats perfection that never arrives
Add roughness to long-term plans; detailed planning fails under high uncertainty

Randomized algorithms solve hard problems faster than deterministic ones
Simulated Annealing: start hot/random, cool gradually toward the solution (escape local maxima)
Add controlled randomness to escape ruts and discover creative solutions

Increase gradually ("a little more, a little more") but cut sharply when you hit limits ("too much")
Apply to career progression, projects, relationships: push until failure, reset to half-effort, rebuild
Prevents burnout and manages others' expectations

Provide constant small acknowledgments ("yeah," "uh-huh") to keep speakers engaged
Your feedback literally shapes what they say; distracted listeners destroy narratives

When something fails, wait exponentially longer before retrying (1 sec, 2 sec, 4 sec)
Offers "finite patience with infinite mercy"—never permanently give up on people or problems

Rational individual choices often produce bad group outcomes (prisoner's dilemma, tragedy of commons)
Add consequences to bad behavior (make contracts binding, enforcement visible) to shift equilibrium toward cooperation
Don't rely on willpower alone

Structure systems where lying doesn't help (truthful bids, transparent incentives)
Eliminates recursive overthinking ("what do they think I think?")

Give concrete options ("Tuesday 2pm or Thursday 3pm") not open-ended choices
State your preferences explicitly; don't ask "what do you want?"—shift burden from group to you

This week: Apply 37% Rule to one upcoming decision (apartment, hire, commitment)
This month: Implement one scheduling/priority change (weighted priority formula or task batching)
Ongoing: Shift explore/exploit balance based on your life stage; audit what you're optimizing for
Relationships: Add one backchannel (active listening) or game-rule change (explicit preferences) to improve communication
Long-term: Accept "good enough" solutions on 50% of decisions; stop perfecting things that don't recur