⭐ Rating: 4.6/5 ★★★★½ — A rare book that makes you smarter about thinking itself
Best for: Professionals overwhelmed by decisions, curious thinkers who love science, anyone who wants a rational framework for everyday life
Reading time: ~7 hours (guide: 14 min summary)
Difficulty to apply: Low to Moderate — concepts click fast; habits take practice
Algorithms to Live By in one minute
Every day you face the same fundamental problems that computer scientists have spent decades solving: when to stop searching for a better option, how to balance trying new things versus sticking with what works, how to organize your space and time, and how to make decisions under uncertainty. Brian Christian and Tom Griffiths argue that the elegant algorithms coded into software are not just engineering tools — they are the distilled wisdom of mathematicians applied to universal human dilemmas. The 37% rule tells you exactly when to stop looking for a house, a job, or a partner. Explore/exploit theory explains why your 20s should look different from your 50s. Least-recently-used caching explains why a messy desk might actually be rational. Read this book and you will never look at your to-do list, your cluttered inbox, or your life choices the same way again.
Key takeaways
- The 37% Rule: When choosing from a sequence of options, spend the first 37% of your time or pool just observing — then commit to the next option that beats everything you have seen so far. This is mathematically provable as the optimal stopping strategy.
- Explore early, exploit later: The famous multi-armed bandit problem shows that exploration is worth most when you have a long time horizon ahead. As your window shrinks — a project deadline, a career phase, even life itself — shift decisively toward exploiting your best-known option.
- Not everything needs sorting: Sorting costs time. Only sort things you plan to search repeatedly. A messy stack of papers organized by recency is often more efficient than an alphabetically perfect filing system.
- Your brain is an LRU cache: Least-recently-used caching keeps what you used last at the top — because what you needed recently is most likely needed again soon. Forgetting is a feature, not a bug; it clears space for new information.
- Schedule by weight-to-deadline ratio: To minimize the total pain of waiting, tackle the shortest task first (SPT rule). When tasks have urgency, weight by deadline. Preempt low-priority work when something critical arrives.
- Bayes updates beat gut feelings: Start with a reasonable prior, then update it as evidence arrives. Most human errors in prediction come from ignoring the base rate — the prior probability before personal data arrives.
- Overfitting is a thinking trap: A model that fits your past data perfectly often predicts the future badly. In life: resist the urge to draw grand conclusions from small samples. The simpler explanation usually generalizes better.
- Relax constraints to find solutions: When a problem feels unsolvable, try relaxing one constraint. Many hard optimization problems become tractable when you allow approximate answers — which are often good enough in practice.
- Randomness is a strategy, not a failure: Monte Carlo simulations, simulated annealing, and randomized algorithms outperform deterministic ones on many real-world problems. Sometimes the best move is a calculated gamble.
- Communication overhead is real: In networking, every acknowledgment and handshake costs bandwidth. In life, excessive check-ins, meetings, and status updates can overwhelm the actual work — a lesson in knowing when to stop signaling and start doing.
What is Algorithms to Live By about?
Algorithms to Live By (2016) by Brian Christian and Tom Griffiths translates eleven foundational computer science algorithms — from optimal stopping to game theory — into practical frameworks for human decision-making, showing how mathematical logic can clarify life’s most persistent dilemmas, from choosing an apartment to managing memory to deciding when to quit.
About the author
Brian Christian is an American author, poet, and researcher whose work sits at the intersection of computer science, cognitive science, and philosophy. His previous book The Most Human Human earned him a place in the Turing Test as a human confederate. Christian studied philosophy, mathematics, and computer science at Brown University and UC Berkeley. His deep collaboration on Algorithms to Live By came through Tom Griffiths, a cognitive scientist at Princeton (now Berkeley) who directs the Computational Cognitive Science Lab. Griffiths uses Bayesian models and probabilistic reasoning to understand how the mind makes decisions. Together, Christian and Griffiths combine literary clarity with rigorous academic grounding — a pairing that makes a genuinely technical subject feel like a long, brilliant conversation over coffee. Explore all Brian Christian book summaries →
Key concepts at a glance
| Concept | What it means | Use it when |
|---|---|---|
| Optimal Stopping (37% Rule) | Observe 37% of options, then pick the next best | Hiring, renting, dating — any sequential search |
| Explore/Exploit | Balance trying new options vs. using known best | Career choices, restaurant picks, project strategy |
| LRU Caching | Keep most recently used items accessible | Organizing desk, bookmarks, mental priority lists |
| Shortest Processing Time (SPT) | Tackle shortest tasks first to minimize total wait | Daily task lists, inbox management |
| Bayesian Updating | Update beliefs proportionally as evidence arrives | Forecasting, evaluating people, forming judgments |
| Overfitting Avoidance | Prefer simpler models that generalize better | Learning from experience, pattern recognition |
| Constraint Relaxation | Loosen one rule to make a hard problem tractable | Negotiation, design problems, planning |
| Randomized Strategies | Introduce deliberate randomness for better outcomes | Breaking routines, brainstorming, sampling |
Part 1: When to Stop Searching — The Mathematics of Optimal Stopping
The first great insight of Algorithms to Live By is that the problem of deciding when to stop looking has a mathematically correct answer. Known in computer science as the “secretary problem,” optimal stopping asks: if you must evaluate candidates one at a time with no going back, how many should you consider before committing to the next great one? The answer — proven rigorously — is exactly 37% of your available pool or time window.
If you have 100 job applicants, interview 37, then hire the next one who surpasses all previous candidates. If you plan to apartment-hunt for 30 days, spend 11 days looking without committing, then take the first one that beats everything you have seen. The math is beautifully indifferent to the domain. Christian and Griffiths walk through the derivation with care, showing how the number 1/e (approximately 0.368) emerges from the calculus of regret — how to minimize the chance of stopping too early or too late.
What is striking is how this simple rule dissolves the anxiety of commitment. The agonizing feeling that there might be something better? That is not a character flaw — it is the natural experience of being in the exploration phase. The rule says: feel that, use it, and then flip the switch at 37%. After that, the algorithm demands decisiveness.
Part 2: The Exploration Dilemma — Exploit What You Know or Discover Something New
The explore/exploit tradeoff is perhaps the most universally applicable concept in the book. Imagine a row of slot machines (the multi-armed bandit problem). Each machine has an unknown payout rate. You can exploit the machine you currently believe is best, or explore a new machine to learn whether it might be better. The optimal balance between these two strategies turns out to depend entirely on your time horizon.
This explains why novelty-seeking is rational in youth and stability-seeking is rational in later career stages — not as a character evolution but as pure mathematics. When you have many pulls left on the machine of life, exploring widely pays off because any new discovery will compound over a long future. When the time horizon shrinks, switching to exploit your best-known option is simply optimal. The algorithms here — epsilon-greedy, Upper Confidence Bound, the Gittins Index — have direct analogies in how we should approach careers, relationships, creative projects, and even restaurant choices.
One counterintuitive insight: the Upper Confidence Bound algorithm says you should be especially willing to try options about which you are uncertain, not just options that look good. Uncertainty is value — it represents potential upside you have not yet measured. This reframes risk-taking: it is not recklessness, it is information gathering.
Part 3: Order and Memory — Sorting, Caching, and Scheduling Your Life
The middle section of the book attacks the ordinary chaos of organized life. Sorting, caching, and scheduling are three of the most studied problems in computer science — and three of the most overlooked in personal productivity advice. The revelation here is that the answers are often counterintuitive.
On sorting: the key insight is that sorting only makes sense if you plan to search the same data repeatedly. A perfectly alphabetized bookshelf is a monument to past labor if you always remember where your books are by recency. The O(n log n) cost of sorting is only justified by the O(log n) savings on repeated searches. If you never search, do not sort.
On caching: your brain already uses an LRU (least-recently-used) strategy. Forgetting is not failure — it is the brain pruning items that, by all evidence, you no longer need. The authors make a compelling argument that a messy desk arranged by recency is a perfectly rational LRU cache. The things on top are the things used most recently; the pile organizes itself. The filing cabinet is the cache miss.
On scheduling: when you have multiple tasks, the Shortest Processing Time rule minimizes the total amount of time people spend waiting across all tasks — a genuinely surprising result. But when tasks have deadlines and weights, you should prioritize by the ratio of importance to time required. Context-switching is expensive in computers and in brains; the authors argue strongly for batching similar tasks and protecting focus blocks.
Part 4: Thinking Better — Bayes, Overfitting, Randomness, and Game Theory
The final section of the book moves into the territory of reasoning itself. Bayes's theorem — update your beliefs proportionally to the evidence — is presented not as a statistical curiosity but as the normative standard for rational thought. The authors show how ignoring the prior (the base rate before your personal experience) is the root cause of most prediction failures in medicine, courts, and everyday judgment.
The overfitting chapter is one of the most practically valuable in the book. In machine learning, overfitting means your model fits historical data so precisely that it fails to predict new data. The cure is regularization: deliberately penalizing complexity, forcing the model toward simpler explanations. The human analogy is profound. The person who draws sweeping life lessons from a single failure, or who designs elaborate routines around one productive week, is overfitting. The wiser move is to assume the simplest explanation consistent with your evidence.
Randomness gets a rehabilitation. Monte Carlo sampling, simulated annealing, and randomized algorithms are not admissions of defeat — they are provably superior to deterministic approaches on many NP-hard problems. The authors argue that spontaneity, creativity breaks, and deliberate randomization in your routines are not irrational; they are algorithmically justified search strategies. The final chapter on game theory shows how the logic of Nash equilibria and dominant strategies applies to negotiations, traffic, and even the social dynamics of email response time.
Who is Algorithms to Live By best for — and who should read something else first?
Perfect for: Knowledge workers drowning in decisions, managers and team leads who want principled frameworks, curious readers with some quantitative appetite, and anyone who has ever agonized over a life choice and wondered if there was a better way to think about it. Also ideal for programmers wanting to see how their craft applies to human life.
Read something else first if: You are looking purely for emotional, narrative-driven wisdom — this book is intellectually rigorous, and readers who prefer pure storytelling may find the algorithmic framing dry. In that case, try Atomic Habits or Thinking, Fast and Slow first. Both pair naturally with this book once you are ready for the computational layer.
Questions to reflect on
- Where in your life are you still in the “explore” phase — and where have you been avoiding the switch to exploit?
- Think of your last major commitment (job, home, relationship). Did you stop searching too early, too late, or close to 37%?
- What does your desk, inbox, or file system actually look like — and is it organized by recency or by some system that costs more to maintain than it saves?
- Where are you overfitting? Where have you drawn sweeping rules from tiny samples in your own life?
- Which of your recurring decisions would benefit most from deliberate randomization — and why have you been avoiding that?
🔥 Ready to let algorithms guide your decisions?
Pick up your copy of Algorithms to Live By and start applying computer science to everyday life.
How to apply Algorithms to Live By (7-day plan)
- Day 1 — Map your open searches: List every decision you are currently in the middle of (job search, apartment hunt, vendor selection). Identify how far into the pool you are — are you past the 37% threshold yet?
- Day 2 — Audit your explore/exploit balance: For one major area of life (career, relationships, hobbies), honestly assess: are you in the right phase? If your time horizon is long, are you exploring enough? If short, are you committing?
- Day 3 — Try LRU on one space: Choose one physical or digital space (desk, bookmarks, inbox folder) and reorganize it purely by recency — most recent at the top, oldest buried. Notice how it feels to use it.
- Day 4 — Apply SPT to your task list: Instead of tackling the most important task first, start your morning by clearing your three shortest tasks. Observe how differently the day feels when small things stop accumulating.
- Day 5 — Practice a Bayesian update: Take one belief you hold about someone or something. Write down your prior (what you believed before evidence). Then write the evidence you have actually seen. Update explicitly, in writing.
- Day 6 — Find where you are overfitting: Identify one rule or routine you built around a single experience or a very small sample. Deliberately simplify it — remove one constraint — and observe whether outcomes change.
- Day 7 — Introduce deliberate randomness: For one repeated decision this week (where to eat, how to solve a problem, who to contact), use randomization to pick. Notice what you discover that your habitual choice would have blocked.
Frequently asked questions
What is the 37% rule in Algorithms to Live By?
The 37% rule (also called the optimal stopping rule) says that when evaluating sequential options with no ability to go back, you should spend the first 37% of your option pool simply observing — without committing. After that threshold, you commit to the very next option that beats everything you saw in the observation phase. Mathematically derived from the number 1/e (≈ 0.368), the rule maximizes your probability of selecting the best overall option. It applies to hiring, apartment hunting, dating, and any sequential search where you cannot revisit passed options.
Is Algorithms to Live By based on real science?
Yes. The book draws directly on peer-reviewed computer science, cognitive science, and mathematics. Tom Griffiths is a Princeton and UC Berkeley cognitive scientist whose lab uses computational models to study human decision-making. Brian Christian conducted extensive primary research with computer scientists and cognitive researchers. The algorithms described — optimal stopping, multi-armed bandits, Bayesian inference, LRU caching, scheduling theory — are established mathematical frameworks, not pop-science metaphors. Some critics note that real-world applications require assumptions that may not always hold, but the core science is solid.
How does the explore/exploit tradeoff apply to careers?
In career terms, the explore/exploit tradeoff suggests that early career stages should involve broad exploration — trying different roles, industries, and approaches — while later stages should shift toward exploiting the skills and networks you have built. This is not just conventional wisdom; it follows from the math of multi-armed bandit algorithms, which show that exploration pays off most when you have a long time horizon ahead (more future pulls on the slot machine). Practically: if you are in your 20s or 30s, trying something new is almost always worth the cost. After 40–50, doubling down on your established strengths tends to compound better.
What does the book say about procrastination and scheduling?
The book reframes procrastination through the lens of scheduling theory. From a computational standpoint, doing the shortest task first (SPT rule) minimizes the total amount of time everyone waits, which is why clearing small tasks before large ones can feel disproportionately satisfying — it actually is optimal by a measurable standard. The authors also note that context-switching is genuinely expensive (computers slow down when switching between processes, as do human brains), lending algorithmic support to deep-focus work blocks. Procrastination on low-priority tasks, when truly important work is present, is sometimes a correct preemptive scheduling decision.
Who are the authors of Algorithms to Live By?
Brian Christian is an author, poet, and researcher known for The Most Human Human and The Alignment Problem. He studied philosophy, mathematics, and computer science at Brown University and UC Berkeley. Tom Griffiths is a professor of cognitive science at Princeton and UC Berkeley who directs the Computational Cognitive Science Lab. His research uses probabilistic and Bayesian models to understand how the mind solves hard problems. Their collaboration brings together Christian's literary clarity and Griffiths's mathematical depth to produce one of the most intellectually rigorous popular science books of the past decade.
What is overfitting and how does it apply to everyday life?
In machine learning, overfitting occurs when a model fits training data so precisely — capturing every noise and outlier — that it fails to predict new data accurately. Christian and Griffiths argue that humans do the same thing: we build life rules from tiny samples, see patterns in random events, and draw elaborate conclusions from single experiences. The computational cure is regularization — penalizing complexity, preferring simpler models. In life, this means: resist sweeping generalizations from one bad date, one difficult boss, or one failed project. Update your beliefs, but proportionally. The simplest explanation consistent with your evidence is usually the most accurate one going forward.
How does Algorithms to Live By compare to Thinking, Fast and Slow?
Both books are about improving human reasoning, but they approach the problem from opposite directions. Kahneman's Thinking, Fast and Slow catalogs the ways human intuition fails — the biases and heuristics that lead us astray. Algorithms to Live By takes a more charitable view: it argues that many apparent human irrationalities are actually rational approximations to hard computational problems. Where Kahneman says “your gut is wrong,” Christian and Griffiths often say “your gut is doing something computationally clever.” The books are deeply complementary — read Kahneman first for the diagnosis, then Christian and Griffiths for the prescriptions.
Related summaries
If you enjoyed Algorithms to Live By, these TGR summaries explore adjacent ideas:
- Co-Intelligence by Ethan Mollick — applying AI as a thought partner, with the explore/exploit framing for working with new tools
- The Alignment Problem by Brian Christian — Christian's follow-up book, going deeper into the ethics and technical challenges of AI
- Prediction Machines — the economics of AI as a decision tool, pairing well with the Bayesian reasoning chapters
- Best AI & Technology Books → — our full ranked list of essential reads in the technology silo