September 18, 2026 - Mindful AI Hacks

Algorithms to Live By: What Computer Science Knows About Your Life That You Don’t

September 18, 2026

Here’s the pitch of this book in one line: the problems you wrestle with every day — when to stop apartment hunting, what to throw away, how to plan a schedule that keeps falling apart — are problems computer scientists solved decades ago. Not for humans. For machines. But the maths doesn’t care which one is asking. Brian Christian, working with cognitive scientist Tom Griffiths, went looking for the overlap between computer science and human life, and found it’s much bigger than anyone expected. Sorting algorithms explain how to organise your kitchen. Caching explains what to keep on your desk. A problem called “optimal stopping” explains dating, hiring, and house hunting all at once. This isn’t a productivity book dressed up in science words. It is a translation of actual math into algorithms to live by with practical examples you can apply to decisions you are making this week. Brian Christian, working with cognitive scientist Tom Griffiths, went looking for the overlap between computer science and human life, and found it’s much bigger than anyone expected. Sorting algorithms explain how to organise your kitchen. Caching explains what to keep on your desk. A problem called “optimal stopping” explains dating, hiring, and house hunting all at once. 🔍 The 37% Rule: when to stop looking and choose This is the single most quoted idea from the book, and it earns the reputation. The setup, in computer science, is called the “secretary problem”: you’re interviewing candidates one at a time, you must decide to accept or reject each one on the spot, and you can never go back to someone you rejected. How do you maximise your odds of picking the best one? If you want to geek out on the actual math behind it, you can read the original definition on Wikipedia’s Secretary Problem page. The proven answer: spend the first 37% of your search purely looking — reject everyone, no matter how good — then commit to the first candidate afterward who beats everyone you’ve seen so far. That number isn’t a rough guess. It’s the mathematically optimal stopping point. It gives you roughly a 37% chance of landing the actual best option — which sounds unimpressive until you realise that’s the best any strategy can do under those rules. This maps almost perfectly onto real decisions: You’re not trying to guarantee the best outcome. You’re trying to stop losing to indecision. The deeper comfort here: regret over “did I look enough” and regret over “did I settle too fast” are both irrational once you know the math. There’s a correct amount of looking, and it’s less than your anxiety tells you. 🗑️When Over-Sorting Costs More Than Mess: Algorithms to Live By Practical Examples Computer scientists obsess over sorting because it’s genuinely costly, computationally speaking. And Christian’s insight is that we over-sort our physical lives for the exact same reason it’s expensive for a computer: most of what you sort, you never look for again. Your bookshelf doesn’t need to be alphabetised if you rarely search it by author. Your filing cabinet doesn’t need a perfect system if you retrieve five documents a year from it. The computer science answer to “should I organise this?” isn’t “yes, always.” It’s a real trade-off: Sorting cost vs. searching cost. If you’ll search something constantly, sort it once, thoroughly. If you’ll search it rarely, don’t bother — a quick scan through mess is cheaper than the hours spent maintaining order nobody uses. This alone is permission to stop feeling guilty about the messy drawer. Some mess is computationally correct. If you’re struggling to let go of perfectionism in your home organisation, check out our previous post on [Why Productivity Isn’t About Perfection] (Internal Link Placeholder). 📥 Caching: keep what you touch, drop what you don’t Related idea, and one of the most immediately useful in the whole book. Computers use “caching” — keeping recently and frequently used data somewhere fast to reach, and pushing rarely used data somewhere slower and out of the way. Not because the rare stuff doesn’t matter. Because retrieval speed should match how often you’ll need it. Apply this literally to your physical space: Most people organise by category (all the pens together, all the cables together) instead of by frequency of use. Christian’s point is that frequency should win. A rarely used “important” document can live further away than a daily-used trivial one — access speed should track usage, not sentimental importance. This is also, quietly, the best argument for a capsule wardrobe you’ll ever read in a computer science book: the clothes you actually wear belong in the fastest-access spot. Everything else is cache eviction waiting to happen. ⚖️ Explore vs. Exploit: the tension behind every choice This is the idea I think about the most, months after reading it. Every time you choose, you’re balancing two competing instincts: Computer science has a precise answer for when to favour which, and it comes down to one variable: how much time do you have left? To understand why this balance is crucial for long-term happiness, we recommend reading The Psychology of Money: A Masterclass in Perspective. Early on — lots of time horizon remaining — exploration wins. Try the new restaurant. Take the unfamiliar project. The potential upside of discovering something great has a long runway to pay off. Later — less time horizon remaining — exploitation wins. Go back to what you know is good. There’s less time left to benefit from a discovery, so the sure thing becomes more valuable than the gamble. This explains something that used to look like a personality flaw and is actually optimal behaviour: why older people order the same dish at the same restaurant, and why younger people are more willing to try the weird new place. It’s not closed-mindedness. It’s correct math for a shorter remaining horizon. It also reframes career and life decisions. A season of trying many things isn’t indecision — it’s the right strategy for

Impromptu by Reid Hoffman: The AI Book That Was Half-Written by AI

Most books about AI are written about AI. This one was written with it. When exploring Impromptu Reid Hoffman AI book insights from 2023, you find a fascinating experiment: LinkedIn co-founder and early OpenAI investor Reid Hoffman sat down with GPT-4 to write an entire book alongside it. Not “AI-assisted” in the vague way everyone claims now. The actual model output is printed right there on the page, replies and all, including the times it was wrong. That single decision is what makes the Impromptu Reid Hoffman AI book worth your time three years later. It’s less a prediction and more a field recording of what it’s actually like to think with a machine. You can explore more about his methodology on the official Reid Hoffman website. Let’s get into the ideas that still hold up — and the ones you should use tomorrow. 🧠 The core argument: amplification, not replacement Hoffman’s central frame is simple, and it’s the antidote to most of the noise online. AI is not here to replace your thinking. It’s here to amplify it. An amplifier doesn’t make music. It makes existing music louder and reach further. Plug a bad guitarist into a great amp and you get loud bad guitar. Plug a good one in and you get a stadium. That’s the honest version of what these tools do. If you bring clear thinking, context and taste, AI multiplies all of it. If you bring nothing and hope the machine supplies the thinking, you get fluent, confident, empty output — and readers can smell it instantly. AI doesn’t give you ideas. It gives your ideas leverage. He also makes a longer historical point: humans have always been tool-shaped creatures. Fire, writing, the printing press, the calculator. Every one of them triggered a panic about what we’d lose, and every one of them expanded what a normal person could do. He sees language models as the next entry on that list, not a break from it. You don’t have to fully agree. But it’s a much more useful starting position than either “this changes nothing” or “this ends everything.” AI doesn’t give you ideas. It gives your ideas leverage. He also makes a longer historical point: humans have always been tool-shaped creatures. Fire, writing, the printing press, the calculator. Every one of them triggered a panic about what we’d lose, and every one of them expanded what a normal person could do. He sees language models as the next entry on that list, not a break from it. You don’t have to fully agree. But it’s a much more useful starting position than either “this changes nothing” or “this ends everything.” ✍️ Copilot, not autopilot The most practical distinction in the book, and the one I’d tattoo on every beginner’s forehead. Autopilot thinking: ask for the thing, copy the output, publish. You are a middleman between a prompt box and a screen. Zero value added, and the work is generic because it was built from the average of everything. Copilot thinking: you stay in the driver’s seat. You supply the direction, the context, the judgement about what’s good. The AI handles volume, structure, first drafts, alternatives, the boring 70%. The difference shows up in the output every single time. Copilot work sounds like a person. Autopilot work sounds like a press release from nowhere. Hoffman’s own workflow in the Impromptu Reid Hoffman AI book is exactly this: he asks, reads critically, pushes back, disagrees, asks again, and keeps the parts where the model failed so you can see the seams. Applying this concept of “leverage” applies beyond just AI. Much like using tools to build wealth, as discussed in our guide on Why Your Net Worth Matters More Than Your Salary, AI is just a tool that requires a human operator to direct it. 🔍 He left the mistakes in — and that’s the best chapter This is the bit I respect most. GPT-4 invents citations in this book. It fumbles arithmetic. It gives confident, plausible, wrong answers. And instead of quietly editing those out to make AI look magical, Hoffman prints them. The lesson for you and me is not “AI is unreliable, don’t use it.” It’s more precise than that: Language models are excellent at shape and unreliable at facts. They’re brilliant at structure, phrasing, brainstorming, reframing, summarising something you already have, and pushing you out of a rut. They’re genuinely risky for statistics, dates, citations, legal specifics, medical claims and anything numerical. So build your workflow around that split. Let AI do the shaping. You do the verifying. Never, ever publish a number or a source you haven’t checked yourself — that one habit will save you the embarrassment that’s caught out lawyers, journalists and a lot of bloggers by now. 🎓 The tutor in everyone’s pocket Hoffman is at his most convincing on education. The old problem: one-on-one tutoring produces dramatically better outcomes than classroom teaching, but there aren’t enough tutors on earth and most families can’t afford one. The new possibility: a patient, infinitely available explainer that never sighs when you ask the same question the fourth time, meets you at whatever level you’re actually at, and can re-explain anything using something you already understand. This is the single highest-value everyday use of AI, and it’s criminally underused. People use these tools to write captions and completely miss that they could be learning compound interest, Excel formulas, or their own kid’s maths homework with them. Try this, right now: Prompt: “Explain [topic] to me like I’m smart but completely new to it. Use an analogy from cooking. Then ask me three questions to check whether I actually understood it.” That last line is the important one. Being quizzed is what turns reading into learning. 💼 Small teams, big output For anyone building something alone — and I know a lot of you are — this is the chapter that matters. Hoffman’s argument in the Impromptu Reid Hoffman AI book