Feynman's Garden |<br>marginalia.nuFeynman's Garden<br>Posted: 2024-05-26<br>The best description of my problem solving process is the Feynman algorithm,<br>which is sometimes presented as a joke where the hidden subtext is “be smart”, but<br>I disagree. The “algorithm” is a surprisingly lucid description of how thinking works in the<br>context of hard problems where the answer can’t simply be looked up or trivially<br>broken down, iterated upon in a bottom-up fashion, or approached with similar methods.<br>Feynman’s thinking algorithm is described like this:<br>Write down the problem<br>Think real hard<br>Write down the solution<br>(This algorithm may or may not have actually been used by Feynman, it’s as far as I can tell<br>Murray Gell-Mann’s account of Feynman’s approach.)<br>The trick is that there is no trick. This is how thinking works.<br>It appears that when you feed your brain related information, without further active involvement,<br>it starts to digest the information you’ve fed it.<br>Thinking is a background process. There’s no Sherlock Holmes-style string of brilliant deductions;<br>the brain is a connection-making machine. If you feed it data points, it will find ways of connecting<br>the dots.<br>So to solve a problem, you start by feeding your mind the relevant information in a fairly literal sense.<br>It’s a direct parallel to how you’d prompt a LLM. Just bring up all the things you think are relevant.<br>Imagine if you were to task someone else with solving the problem, imagine what they might have use of knowing,<br>really go over anything you think might be useful. It’s good to write that down in the simplest way possible.<br>It should be very clear to you what question you are attempting to answer. What is written on paper is not the<br>important thing, the act of assembling the question is the work!<br>To answer the question, what you do next is to remove yourself from further inputs.<br>As long as you keep feeding your brain with more data, the queries never resolve, and<br>eventually the stuff you fed it first will fall out of the context window and amount to nothing.<br>For this reason, it’s crucial to step away from the keyboard. Go for a walk, or sit down in a quiet place,<br>take a shower; whatever you choose to do, just keep thinking about the problem.<br>As you do this, all the background processing you kicked off earlier will begin to resolve.<br>Not immediately, it may take a while—hours, sometimes days—but as it does you get sudden<br>insights into the problem domain. You may or may not immediately find a solution to the problem,<br>but you’ll often at least find a deeper understanding.<br>The conclusions must be written down. Not only because these insights are fleeting and<br>can be hard to recall, it also appears to act as a sort of feedback to the brain that the<br>information is important, that it’s on the right track, to keep going.<br>The perhaps most inspired period of my life was sometime in 2021, before I’d moved<br>in with my girlfriend. What I’d inadvertently ended up doing was to arrange my life<br>into something very similar to the algorithm above.<br>I’d grind away and work on the search engine all week, filling my brain with all manner of information<br>related to the work I was doing, and then I’d go visit her over the weekend, not bringing a computer.<br>Without exception, by the time the weekend was over, I’d have several pages worth of new ideas written<br>down in a notebook that I’d implement over the following week, rinse and repeat.<br>I keep coming back to the gardening metaphor of thought.<br>What you think today is the result of what you read yesterday. If you want to have relevant,<br>interesting, useful thoughts; use your reading as a template. It will not only model the topics<br>you are thinking about, but the very thoughts themselves. It’s as much pruning gossip and<br>irrelevant noise as it is planting the interesting and insightful and well thought-out.