Algorithms to Live By
Computer science for everyday life: optimal stopping, sorting, caching and Bayesian thinking applied to human decisions.
Want the full book?
The summary saves time, but the original gives you depth.
The premise
Computer scientists have spent decades working out how to make optimal decisions under constraints. The authors argue those solutions apply directly to human problems: when to stop looking, what to keep, how to schedule, how to update beliefs.
Optimal stopping
The secretary problem gives a genuinely surprising answer: to maximise your chance of picking the best option, look at roughly the first 37 percent of candidates without committing, then take the next one better than everything you have seen. It reframes when to stop searching.
Explore versus exploit
Every choice trades gathering information against using what you already know. The mathematics of the multi-armed bandit gives a principled answer: explore more when you are young or the stakes are small, and exploit more as the window closes.
Sorting and keeping things
Sorting is about what to do next; caching is about what to keep. The optimal approach depends on how often you use something compared with how expensive it is to retrieve — which is why your desk, your inbox and your computer memory should not all be organised the same way.
Scheduling
The book explains why shortest-job-first is optimal when tasks are independent, why interruption has a real cost, and why priorities must be explicit. Deadline pressure leaks into everything else you are doing.
Bayes's rule
Predictions should be updated as evidence arrives, weighted by how likely that evidence was under each hypothesis. The authors show how failing to update — or over-updating on weak evidence — produces bad judgements.
Overfitting
A model that fits the past perfectly often predicts the future badly. This is the mathematical version of "too clever by half", and it applies to plans and theories as much as to data.
Relaxation and randomness
Hard problems often become tractable if you relax a constraint, and randomness is a tool rather than a failure — random strategies can be optimal and can defeat predictability.
Computational kindness
The closing chapter turns the lens outward: respecting other people's attention and cognitive limits is a computational problem too, and solving it makes you easier to live and work with.
🔑 Key actionable ideas
- Look at about 37 percent of options, then commit to the next best.
- Explore when the cost of being wrong is low; exploit when it is high.
- Keep things according to how often you actually need them.
- Do shortest jobs first, and treat interruption as a real cost.
- Update your beliefs in proportion to the evidence.
- Do not overfit your plans to the past.
- Respect other people's limited attention as a design problem.
Who is this book for?
Anyone who enjoys ideas and wants better mental tools for decisions, time and attention.
Final verdict
A genuinely useful crossover between computer science and daily life. Readable, practical and full of surprising answers.
