Central argument
Christian, a writer, and Griffiths, a cognitive scientist, take the problems computer science has worked on for decades (searching, sorting, caching, scheduling, predicting) and show that people face the same ones every day. Choosing a flat, deciding whether to go back to a favourite restaurant, keeping a desk in order, filling a to-do list: each of these has a well-studied algorithmic answer. The point is not to turn life into code. Once you accept that you will never have all the information or all the time, some strategies are simply better than others, and many habits we feel guilty about (the messy pile, giving up the search, going back to the familiar) turn out to be rational. A good decision is judged by the process, not by how things went.
Examples / Notable passages
- The 37% rule. Commit too early and you miss better options; wait too long and the best ones are gone. The optimal strategy (the "secretary problem") is to spend the first 37% of the time, or of the candidates, only looking, then take the first option that beats everything seen so far.
- New thing or old favourite. The explore/exploit trade-off depends on how much time is left, not on mood. A new restaurant pays off only if there is time to go back once you've found it: with a lot of time ahead you explore, near the end you exploit what you already trust. The book uses this to read the way older people narrow their circle to a few close friends as a sensible choice, not a loss.
- The self-organizing desk. Filing everything into rigid folders costs a lot and gives little back. A plain pile, where whatever you used last goes back on top, sorts itself by recency: what you use often stays at the top, what you never touch sinks. It's the same logic computer memory uses (keep the most recently used items closest at hand).
- Busy is not the same as done. Clearing small tasks one after another feels productive while the important projects stay untouched. Scheduling only makes sense once you've decided what you're trying to optimize: finish what matters, not tick off as many boxes as possible.
- You get what you measure. Turn a complex goal into a single number and people start optimizing the number instead of the goal. The metric becomes a proxy that rewards rushing and cutting corners. Before chasing a target, it's worth asking what it's standing in for.
- Your feed warps your reality. We estimate what is common or dangerous from what we've been exposed to. News and feeds select rare, dramatic events, so the sample is skewed and our sense of risk with it. Good judgment starts with checking the inputs.
- A step back beats a dead end. A plateau is often a local maximum: every nearby move looks worse, so you stay. Algorithms get out by accepting a temporary step back, or by adding a bit of randomness, to reach a higher peak further on.
- Computational kindness. "Whatever you want" sounds polite, but it hands the other person the whole job of guessing everyone's preferences. Saying what you'd like, or offering two or three options, is kinder because it takes work off them.
Objections
- The book works best where the problem can be defined: a fixed number of options, a clear goal. Many real choices (a vocation, a relationship) don't have a known "pool" to sample 37% of, and the clean numbers can give false confidence.
- Reading habits as algorithms can turn into a justification for them. The messy pile is efficient for papers; "it's rational" does not prove that every mess or every shortcut is.