Algorithms to live by

The book in one line#

Algorithms to Live By by Brian Christian and Tom Griffiths (2016) takes everyday life problems and shows they are problems computer science has already studied carefully. The main spirit is this: you can't make perfect decisions, but you can make them wisely. Some problems have a best answer. For others, "good enough" really is the best possible. So:

The rule

Judge yourself by your process, not by the outcome. A good decision can still turn out badly.

If Four Thousand Weeks is the philosophy of limited time (accept it), this book is the math (how to choose well inside it).

One idea per chapter#

Chapter	Life question	Lesson
Optimal stopping	When do I stop looking (for a flat, a partner, a hire)?	The 37% rule: spend the first 37% of your search only looking. Then take the first option that beats everything you've seen.
Explore vs exploit	Try something new, or stick with what works?	Explore when you have lots of time left; exploit when you don't. Young people try everything; older people keep fewer, closer friends. It's rational, not decline.
Sorting	Should I organize everything?	Sorting is expensive. Don't sort what you won't search.
Caching	What do I keep close, what do I throw away?	Keep what you used most recently at hand. A pile on your desk with the latest on top is a smart system.
Scheduling	What do I do first?	Pick a rule for your goal: nearest deadline first, or shortest task first. Switching tasks has a cost, so batch email and messages.
Bayes's rule	How do I predict from little information?	Start from sensible expectations, then update as information arrives. The news distorts expectations because it reports what's rare.
Overfitting	Should I think harder?	Not always. Overthinking fits the noise. Under uncertainty, simple rules often win.
Relaxation	The problem is too hard.	Solve an easier version first. Drop a constraint ("what if money weren't an issue?"), then work back to reality.
Randomness	I'm stuck.	A bit of randomness escapes a "local best". Travel, meet new people, try something random.
Networking	How do I handle rejection?	Grow slowly, cut back fast when overloaded, like the internet's TCP protocol. After repeated failures, wait longer before retrying instead of quitting.
Game theory	Why do good people get bad results together?	Sometimes the problem is the rules of the game. "Unlimited vacation" makes people take less. Change the game, not just your move.

The ending: computational kindness#

Make decisions easier for other people. Instead of "Whenever works for you!", say "How about Tuesday at 2pm?" An open question hands the hard thinking to the other person. Narrowing their options is a kindness.

Explore vs exploit, in practice#

This is the chapter used most in the rest of the course, because it answers "should I keep looking or commit?"

Time left decides the mix. With lots of time ahead, explore more. As a deadline nears, stick with what works.
Be optimistic about the unknown. If an option is cheap to try and you know little about it, favor it. Its upside is still open.
Big jumps early, small adjustments late. If you're stuck on a small hill, restart somewhere else.
The AI-era twist: each try is now much cheaper; a prototype takes days. When testing is cheap, test more and analyze less.

Scheduling your own work

A weighted list of tasks (Chapter 4) is a scheduling problem. The book's rule for it: sort by importance ÷ time it takes, and do the highest ratio first.

Takeaways

Good process, not good luck, is what you control.
Explore early, exploit late. How far away your deadline is tells you which.
Simple rules beat over-analysis when the future is noisy.
When stuck, relax a constraint or add randomness.