Build it yourself: why making and reinventing beats consuming knowledge

The Short Version

Five people who work on very different problems keep pointing at the same pattern: what you make, rebuild or work out for yourself stays with you, and what you only take in mostly does not [1][2][3][4][5]. Grant Sanderson says the things he has learned best all involved some rediscovery, and warns that even an excellent lecture can be satisfying without sticking [1]. Neil Gershenfeld's fab labs, Joe Liemandt's push to raise builders, Bret Weinstein's distinction between consuming and discovering, and Mary Helen Immordino-Yang's teacher who invents knowledge in front of the class are versions of the same idea in a workshop, a school, the physical world and a lecture hall [2][3][4][5]. The test it gives a learner is simple, though the sources also show where it needs care: making works best with guidance, with the basics already in place, and with a reason to care [1][3][5].

Table of contents

What They Said

Grant Sanderson: reinvent a few things, and distrust what merely feels clear

Sanderson, the math educator behind 3Blue1Brown, starts from Richard Feynman. What he most admires is Feynman's habit of reinventing ideas for himself: when reading a paper, working out what problem it was trying to solve, attempting it himself to gain personal ownership, and only then seeing what others did [1]. Sanderson says the things he has learned best are the ones with some element of rediscovery [1]. Choosing a few core building blocks to approach on your own fills your head with counterexamples and intuitive examples you would never get from just reading a proof [1].

He is honest about the price. Rediscovery slows you down, and he calls it partly a fault, one reason he is not an active researcher [1].

The other half of his argument is a warning about the opposite habit. He describes the "Feynman effect": while watching or reading Feynman's lectures everything makes perfect sense, yet a week later you cannot recall what gave you the insight [1]. He worries his own videos risk being "intellectual candy" in the same way [1]. Unless learners do something active, such as reinventing the idea, working problems or using spaced repetition to lock in the terminology, it will not stick [1]. His message to educators is that how satisfying content feels does not correlate with long-term learning [1].

He applies the same test to mathematics itself. Pattern recognition, he says, "is most of understanding," but in math the patterns are stepping stones that have to be tested [1]. His example is dividing a circle with chords, where the counts run 1, 2, 4, 8, 16 and then 31, not 32 [1]. You only find that out by doing it.

Neil Gershenfeld: give people the tools and they make things nobody assigned

Gershenfeld, who directs MIT's Center for Bits and Atoms, has the most personal stake in the split between thinking and making. He wanted to attend vocational school to learn welding and fixing cars but was told he was too smart, and at Bell Labs he got union grievances for trying to make things in the workshop [2]. He traces the prejudice to the Renaissance division between the liberal arts and the so-called illiberal arts, and argues that micromachining or embedded coding is as expressive as painting or writing a sonnet [2].

To teach students to use his lab's machines he started a class called How to Make Almost Anything, which became one of MIT's most oversubscribed courses [2]. What surprised him was what students made: a device that saves your screams and plays them back later, a dress with sensors and spines that defends personal space, a web browser for parrots, an alarm clock you must wrestle [2]. Of the student who built the scream device he says it was not for a research paper or a business model: "She wanted, it was because she wanted one" [2]. "It was personal expression," he says, and he concluded that the killer app of digital fabrication is personal fabrication [2].

That class grew into the fab lab network, now about 2,500 labs in 125 countries and doubling roughly every year and a half [2]. Seymour Papert, who studied with Piaget, told him fab labs fulfilled his goal of letting kids create robots rather than just program them [2]. In Arctic hamlets, South African townships and rural Indian villages Gershenfeld found the same kind of bright, inventive people who might end up at MIT, a density of underused minds he calls "the greatest natural resource of the planet" [2]. Wherever a lab opens, he says, "they come out of the woodwork" [2]. The promise, in his words, is that "you can make many of the things you consume" [2].

His view of how discovery happens is also a maker's view. He says "essentially, everything I've ever worked on has failed" in ways that led to something better [2]. Instead of ready, aim, fire, he practices ready, fire, aim: do the homework, try something without over-planning, then look carefully at where you actually hit [2]. A failed attempt at low-cost shoplifting tags led to early quantum computations; an experiment ruined by bubbles led to logic built from bubbles [2]. Part of the fab lab network's value, he argues, is giving more people the opportunity to fail [2]. And he notices the same split among students who approach him: 99 in 100 tell him his work is fascinating, while the rare one points out a mistake and, as he puts it, "They just sort of say I'm here and get to work" [2].

Joe Liemandt: school was built to produce consumers

Liemandt, the principal of Alpha School, puts the case as a design flaw in schooling. "Our education system was built to create consumers," he says: a hundred years ago "we lined you up in a classroom and we told you to sit there and listen and write stuff down" [3]. His alternative is stated just as plainly: "We need to create builders. That is what kids need to be doing. That's what adults need to be doing. You can't just be thinking" [3].

At Alpha, students spend about two hours a day on individualized academics and the rest of the day building [3]. His examples include fifth and sixth graders organizing a music festival and raising $25,000 in sponsorships, middle schoolers running vending machines and food trucks, a student producing a Broadway-style musical and a student who built a bike park [3]. One program has students track a 1,000-hour project to show them the gap between their vision and their habits [3]. Building, he says, whether it is climbing, drones and robots, a business, books or videos, is "what engages kids" [3].

His clearest story is a voluntary weekend hackathon on an AI coding tool. Students came in at six on a Friday and worked through the weekend [3]. One group of girls, who he says would not normally do programming, built a tool for chatting together online while watching a reality show [3]. They were building a project they liked, but they came away with the belief "okay, I can be a builder" [3]. "Kids naturally want to build stuff," he says [3]. He reports that 85% of Alpha parents said they saw their child do something they thought was impossible for their age; that figure is Alpha's own, and the episode notes independent trials are still underway [3].

Bret Weinstein: the physical world tells you whether you understood

Weinstein, an evolutionary biologist and former professor, supplies the sharpest statement of the contrast. When Joe Rogan says today's kids have unprecedented access to inspiring ideas online, Weinstein is skeptical [4]. What makes people smart, he argues, is interaction with the physical world that rewards figuring something out, like building a treehouse that tests your understanding [4]. Screens risk turning everything "into an exercise in consuming information rather than discovering" [4].

He sees the same shift across life: we have converted participatory activities into consumables, so people watch sports instead of playing and listen to music instead of singing and playing [4]. Even viral clips of superhuman stunts can be demotivating, because viewers see no pathway to doing it themselves [4].

His classroom practice followed from this. He and his wife, Heather Heying, ran a learn-a-skill exercise in which students picked any objectively measurable skill, not to master it but to see how their own minds learn [4]. Students consistently underestimated the effort required [4]. He believed "almost the entire job of education is about incentives": if a student wants to understand something, most of the work is done [4]. When Rogan pushes back that discipline still matters, Weinstein agrees and calls wisdom "effectively delayed gratification" [4].

Mary Helen Immordino-Yang: the best teachers rebuild the idea while you watch

Immordino-Yang, a professor of education, psychology and neuroscience, brings the pattern to the teacher's side of the room. Andrew Huberman describes the best instructors in a neuroscience course he knows: world experts who teach as if they were, in his words, "learning it again for the first time" [5]. Immordino-Yang names what is happening. Such a teacher is "not a squirrel with nuts and giving all the nuts to the kids" [5]. "He's inventing the knowledge in front of them" [5].

She sets this against what she sees as the usual idea of a good student, someone who acquires a fact and then owns it forever, as in "I know how to do algebra" on demand [5]. She would rather people engage with their own knowledge in curious, flexible ways and learn to "rethink what we think we know" [5].

Her evidence from schools is a group of New York public schools that are exempt from the state exams and instead assess students through months-long interdisciplinary projects in which students must invent the question as well as the work [5]. One student, an immigrant who had never passed a math class, spent months on Zeno's paradox of walking halfway to the door and learned fractions because he needed them [5]. Her own teaching went the same way: a student's question about why early hominids in a documentary always looked dark-skinned grew into a months-long unit on biology, culture and identity [5]. Her path was a maker's path too, through cabinetmaking and documenting traditional boat-building before she taught seventh-grade science [5].

She also gives a small example of producing as a way of understanding. When her young son was bothered by a classroom behavior chart, she had him write his teacher a letter explaining why [5]. Writing it meant formulating what the chart was and why it bothered him, and, she says, "in so doing it helps him not be so bothered by it" [5].

Where They Agree, and Where They Don't

The agreement is broad. All five rank doing above receiving. Sanderson's rediscovery, Gershenfeld's personal fabrication, Liemandt's builders, Weinstein's treehouse and Immordino-Yang's projects are the same move [1][2][3][4][5]. Several of them also trace the problem to the same place. Liemandt and Rogan both describe schools as built to produce obedient, rule-following people rather than makers [3][4]. Gershenfeld and Liemandt both point to the loss of shop class and vocational learning [2][3]. And three of them, Weinstein, Liemandt and Immordino-Yang, say that wanting to know is most of the battle [3][4][5].

But the sources disagree on important details, and a reader should not paper over them.

Should you discover the basics yourself, or be taught them? This is the sharpest split. Liemandt rejects discovery as a way to teach fundamentals. No one, he says, would put a child on a sports team that announced, "We just let you productively struggle and figure it out yourself" [3]. An athlete wants "Give me the step by step" [3]. He notes that inquiry learning versus direct instruction is a fierce debate in education, and he comes down on the side of direct instruction and mastery for academics [3]. Immordino-Yang's project schools, where students invent the question, sit on the other side [5]. The two are less opposed than they look, because Liemandt splits the day: fast, direct, individualized instruction for the basics, then hours of building [3]. His claim is about how to acquire fundamentals; hers is about what makes knowledge meaningful. Still, they would design a math class differently.

How much open exploration? Sanderson and Gershenfeld both hedge. Sanderson notes that friends who build interactive explanations find only a small sliver of users actually experiment, and that a sandbox with five sliders and no guidance can be daunting, so he plays with the simulation himself and then chooses which stories to tell [1]. He recommends reinventing only a few core ideas, not everything [1]. Gershenfeld's ready, fire, aim keeps the first step, doing the homework, and he criticizes fail-fast thinking for dropping the ready and aim parts [2].

Are screens the problem? Weinstein treats screens as a route to consuming rather than discovering [4]. Rogan, in the same conversation, says "I've been educated almost entirely outside of schools," through books, audiobooks, podcasts and conversation [4]. Huberman, talking with Liemandt, says he does not believe you can get all the education you need from YouTube, and that how a person forages through technology is critical [3]. Sanderson wants a small number of excellent explanations online so classroom time can go to everything in teaching that is not explanation, a goal he calls "commoditizing explanation" [1]. So the dispute is not really about the medium. Sanderson's own position reconciles it: a video is fine as an explanation, and useless as the whole of learning [1].

Rewards. Liemandt has paid students for perfect scores on state tests and believes paying for a breakthrough can create intrinsic motivation afterward, though he acknowledges it is controversial and that some parents object [3]. Immordino-Yang argues that "whatever you're having emotion about is what you're thinking about," so when students' emotions center on grades they learn about performance rather than ideas [5]. On her account, a cash prize for a test score risks teaching the wrong lesson; on his, it removes a block such as believing you are not a math person [3][5].

Video games. Rogan cites a finding that surgeons who play them make about 25% fewer errors; Weinstein grants that games train the mind brilliantly but says they do so for skills that quickly become obsolete [4].

Why It Works

Reinventing builds the surrounding map. Sanderson's explanation is that working on an idea yourself fills your head with counterexamples and intuitive examples [1]. A proof read passively gives you the route; attempting it yourself shows you the wrong turns too.

Feeling clear is not the same as learning. The Feynman effect names the gap: a lecture can make perfect sense and leave nothing a week later [1]. Sanderson's point that the satisfaction of content does not correlate with long-term learning explains why consuming feels more productive than it is [1].

Reality grades the work. Weinstein's treehouse tests your understanding [4]. Gershenfeld's failed tags and ruined experiments told him something no plan could, which is why he looks carefully at where a shot actually lands [2]. Sanderson's chord pattern breaks at 31 [1]. Weinstein's students learned how much effort a skill takes only by trying one [4]. In each case the thing you make can fail, and the failure is information.

You learn what you feel something about. Immordino-Yang's principle is that emotion decides what gets learned [5]. Students whose emotions are attached to ideas learn ideas; the student working on Zeno's paradox learned fractions because he needed them for a question he cared about [5]. Huberman suggests a teacher who relearns the material in front of the class ignites the emotional systems of the learner's brain in a way that hearing an expert relay facts does not [5]. Liemandt sees the same from the school side: building is what engages kids, enough that they give up a weekend for it [3]. Gershenfeld's students built what they personally wanted [2].

Making changes what you believe you can do. Liemandt's hackathon students left thinking of themselves as builders [3]. Weinstein describes the reverse effect from watching: with no visible pathway from the clip to the skill, viewers are discouraged rather than inspired [4].

Putting it into words forces understanding. Sanderson withholds a topic from his list until he understands it well enough to explain, ideally with a unifying insight [1]. Immordino-Yang's letter exercise works the same way: to write why something bothers you, you have to work out what it is [5].

The sources give mostly experience and argument for these mechanisms rather than controlled studies. Sanderson speaks from his own learning and from what colleagues see in their analytics [1]; Gershenfeld, Liemandt and Weinstein from their labs, school and classes [2][3][4]; Immordino-Yang from the project schools and her own classroom, alongside her research on emotion [5].

Where It Applies, and Where It Doesn't

The test, making or only taking in, applies to any study method, and it is most useful as a check on methods that feel good. If a lecture, video or book leaves you with nothing you produced, Sanderson's warning applies [1].

It does not mean every fact should be rediscovered. Sanderson limits reinvention to a few core building blocks and admits the cost in speed [1]. Liemandt's argument about fundamentals is the stronger limit. He describes knowledge as layered, with chemistry depending on algebra, algebra on fractions and fractions on math facts, and says anything memorized to fluency frees working memory for harder problems [3]. His example is a student stuck in the low 700s on the SAT math section who scored 790 after going back to memorize her multiplication tables [3]. Fluency of that kind comes from instruction and repetition, not invention. He also names mastery, worked examples and spaced repetition among the things that worked at Alpha [3].

It does not mean unguided play. A sandbox without a story is daunting, and most people do not experiment on their own [1]. Gershenfeld's labs come with a class, a network and, in a newer version, a lab in a box that contains the knowledge of how to use the tools [2].

It depends on access and setting. A fab lab costs about as much as a minicomputer once did, and Gershenfeld says getting started can seem like a big threshold [2]. Sanderson says the thing he most wants to change in his own life is working without colleagues to bump into and share puzzles with [1]. Immordino-Yang adds that people can only think together when they feel safe to [5]. Huberman voices a worry some people have on hearing all this, the feeling of "I can't build things," and Liemandt's answer is to start from what the person already likes [3].

And it has costs in volume. A learner who reinvents will cover less ground than one who reads. Sanderson accepts that trade for a few ideas and describes himself as someone who knows many things rather than one thing deeply [1]. He is no enemy of reading either: his cure for writer's block is to read, anything at all, for 15 to 30 minutes [1]. Consuming is not the enemy; consuming with nothing made at the end is.

What To Do With It

1. Apply the test to how you study now. For each method you use, ask what you produced: a solved problem, an explanation, an object, a question. If the answer is nothing, treat the time as entertainment rather than learning [1].
2. Before reading a solution, try the problem. Work out what the paper, proof or chapter is trying to solve, attempt it yourself, and only then see what others did [1]. Do this for a few core ideas in a subject, not for all of them [1].
3. Follow every lecture or video with something active. Work problems, try to rebuild the idea from memory, or use spaced repetition for the terminology [1].
4. Test the patterns you think you see. Carry the example one step further than the explanation did; the chord count that looks like doubling stops doubling at the sixth term [1].
5. Get the basics fluent by direct means. Where a foundation is missing, go back and fill it with step-by-step instruction and repetition before building on top of it [3].
6. Pick a skill you can measure and learn it to watch yourself learn. Weinstein's exercise is not about mastery but about seeing how your own mind learns, and expect it to take more effort than you think [4].
7. Build something you personally want. A treehouse, a tool, a business, a video, a piece of software: choose a project that can fail in a way you will notice, and start from your own interests [2][3][4].
8. Use ready, fire, aim. Do the homework, try something without over-planning, then study carefully where you actually landed and follow what the failure shows you [2].
9. If you teach, rebuild the idea in front of your students. Work it out as if for the first time instead of handing over the finished result, and let students invent some of the questions [5].
10. Explain it to someone. Write the letter, make the short explanation, or teach the topic; Sanderson's advice to hesitant experts is to publish and see how it goes [1][5].

Sources

1. Grant Sanderson (3Blue1Brown) on Feynman, Exponential Growth, Teaching Online, GPT-3 and Manim · Lex Fridman Podcast
2. Neil Gershenfeld on Self-Replicating Robots, Fab Labs and the Future of Digital Fabrication · Lex Fridman Podcast
3. Joe Liemandt on Alpha School, AI Tutoring, and Learning Faster Through Mastery and Motivation · Huberman Lab
4. Bret Weinstein · The Joe Rogan Experience
5. Dr. Mary Helen Immordino-Yang: How Emotions & Social Factors Impact Learning · Huberman Lab