Be a Source, Not a Relay: Grant Sanderson on Creative Work That Lasts

Life of Luba · Grant Sanderson (@3Blue1Brown) · Video ID: Rtkac4WHC1o

Who Is Grant Sanderson?

Grant Sanderson is the creator of 3Blue1Brown, one of the internet’s most respected mathematics channels. A Stanford math graduate, former Khan Academy educator, and self-taught animator, he built the open-source animation engine behind his visual explanations and turned abstract mathematics into something millions of people can feel, not just recite.

His credibility comes from an unusual combination: mathematical taste, teaching empathy, craft obsession, and a refusal to reduce education to information transfer. His work is not simply “clear explanations.” It is a long-running experiment in how intuition, beauty, motivation, and originality interact.

The Central Thesis: Make People Feel the Shape of an Idea

Sanderson’s best work starts from a deceptively simple premise: the first version of an idea is rarely the clearest version. In research, discovery often happens in fog. The first proof, paper, or explanation may be correct, but it is not necessarily the version that makes the truth feel inevitable. Sanderson is drawn to what he calls “unsolved expository problems”: ideas that are known to be true but still lack the right perspective, the one that makes them feel obvious.

That distinction explains why he never felt fully pulled into academia despite loving mathematics. Academia was the legitimized path for spending one’s life with math, but his actual energy came from outreach, teaching, and re-seeing established ideas. He is less motivated by being first to understand something and more motivated by finding the explanation that turns significance into intuition.

That stance also explains his durability as a creator. Many long-time creators feel trapped by the treadmill: analytics, deadlines, audience expectations, and the need to keep feeding a system. Sanderson still feels energized after more than a decade because the work remains tied to craft and curiosity, not merely output.

From Relay to Source

Sanderson draws a useful line between being a relay and being a source. A relay reads a book, paper, or post and repackages the same point in another format. A source digests the idea until the explanation has its own center of gravity. The topic itself does not always need to be novel; the way of seeing it does.

“To what extent are you a source and to what extent are you a relay?”

This is the heart of 3Blue1Brown’s appeal. His famous linear algebra series did not succeed because linear algebra was an untouched subject. It succeeded because students encountered a version that made the subject click. Years later, university students still thank him for that series because it gave them an intuition they had not received elsewhere.

The lesson is subtle but important: novelty at the topic level is overrated, but originality at the perspective level is non-negotiable. An audience will forgive a familiar subject if the creator gives them a sharper mental model. They will not stay loyal to someone who merely re-broadcasts the current consensus.

The Wedding Speech as a Miniature of the Whole Craft

The conversation begins with an unlikely case study: a wedding speech Sanderson wrote with roughly 24 hours’ notice. The speech apparently wove together nuclear fusion, specific observations about the bride and groom, emotion, timing, jokes, and a scientific metaphor that landed as heartfelt rather than gimmicky.

His process was revealing. He did not ask an LLM to “write a speech,” because he does not think LLMs are good writers in that sense. He used it more like a thinking partner to generate raw material, then built the structure himself. The strongest move was finding the bookends: begin with an overly sciency premise that fit the groom, end by turning that premise into a sincere emotional payoff, and fill the middle with specific anecdotes that show why the couple works.

That is also how strong explanatory work functions. The outer frame creates momentum. The middle earns trust through specifics. The ending makes the structure feel inevitable.

His public speaking advice follows the same principle. When you forget what comes next, do not announce that you are lost. Look meaningfully at one person, hold the silence, and let the pause look intentional while your mind catches up. Pauses make ideas land; they also buy time. Performance is not fake confidence so much as designing the conditions under which attention can survive.

The Algorithm Is Usually Just the Audience in Disguise

Sanderson is not immune to analytics. He admits that watching real-time views is a kind of dopamine hit, even when it teaches him almost nothing. But he rejects the mythology of “the algorithm” as a mysterious sea monster with arbitrary tastes. In most creator conversations, replacing “the algorithm” with “the audience” makes the sentence more accurate.

“If you replace the words ‘the algorithm’ with ‘the audience,’ it’s almost always the same.”

Good thumbnails matter because people need a reason to click. Strong openings matter because humans decide quickly whether they are in good hands. Runtime, title, channel name, and prior trust all shape behavior because viewers are making predictions about whether the experience will be worth their time.

He still thinks metrics can be useful, but only when they reflect the work’s real purpose. Views are a crude metric. Watch time is better. Better still might be something like watch time per month five years from now: a measure of whether the work continues to matter. That metric aligns with his stated aim of making explanations that endure, not merely ones that spike.

Fun Work, Strategic Work, and the Capacity Problem

Sanderson describes the enjoyable stretch of making a project as roughly the middle: after the uncertain first 20% and before the exhausting final 10%. He loves the craft of animating ideas, especially pure math topics that have an aesthetic payoff. His example is the hairy ball theorem: the statement that a hair-covered sphere cannot be combed perfectly flat without a cowlick, while a donut can be smoothed elegantly. It is silly on the surface and serious underneath, which makes it ideal territory for visual explanation.

But the work he most enjoys is not always the work he thinks would be most valuable. He repeatedly returns to the tension between one-off beautiful explorations and curricular foundations. A sequel to his linear algebra series might help a large number of students for years, but those long, foundational projects are easy to defer when a more immediately exciting topic appears.

His answer is not to choose one principle forever. It is to build capacity so the trade-off becomes less binding. A mature 3Blue1Brown could have something like departments: one side creating foundational visual lessons for important topics, another side producing joyful, curiosity-driven explorations that keep the channel alive as a source of wonder.

Write Until the Thought Exists

When Sanderson needs to think through decisions—whether to hire, how to monetize, whether to focus more on machine learning, whether to return to foundational curriculum—he writes. Not polished essays. Not necessarily notes he intends to revisit. Often just daily files whose purpose is to externalize rumination.

“It wasn’t actually a thought… until you were forced to put it into words.”

That line captures a practical epistemology. Writing does not merely clarify thought; it creates thought. Vague mental motion becomes inspectable only when it is forced into language. The point of the note is not archival perfection. The point is to convert “fake thinking” into actual thinking, then use that clarity to make concrete plans.

His system is intentionally imperfect: files, Apple Notes, linked notes, possible curiosity about Obsidian and Notion. The tool matters less than the habit of forcing the unresolved question into words.

Why the Solo Creator Is Building a Team

For years, Sanderson leaned into solo creation. He avoided sponsorships, relied on Patreon, and kept the whole process tightly integrated: writing, animating, editing, and teaching were intertwined. He could not easily hand off “editing” because the animation timeline was often part of how he discovered the explanation.

Now he is rethinking that. The constraint is no longer whether the channel can support him personally. It is whether his own time is the bottleneck preventing the work from becoming what it could be. He can imagine hiring animators, illustrators, and editors; building a more formal structure for collaboration; and spending more of his day learning, writing, and directing highly leveraged work.

The business model matters because the wrong monetization structure would distort the work. Traditional sponsorships optimize for the next few months of views. His proposed alternative is closer to a trusted talent marketplace or “virtual career fair” for mathematically and technically strong people: partner companies he respects, presented as a durable resource rather than a disposable ad read. If the body of work keeps sending the right people to a shared piece of real estate over years, the incentive aligns with long-lived trust rather than per-upload pressure.

He is clear-eyed about the transition. Hiring and onboarding will create a local minimum before it creates a higher local maximum. The short run may mean less learning and less making. The long run could mean more ambitious work without turning the channel into a treadmill.

Education Has an Explanation Problem Only After It Has a Motivation Problem

Sanderson is skeptical of every generation’s claim that a new medium will revolutionize education by improving access to explanation. Motion pictures, radio universities, TV universities, dot-edu startups, MOOCs, and now LLMs all improve the delivery layer. LLMs are genuinely stronger than previous media, but the central bottleneck remains the same: do students want to learn?

His deliberately provocative “solution” is to hire one actor per student, have each actor charismatically befriend a student, and then show authentic interest in the subject the student should learn. The joke works because the mechanism is real. Motivation is social. The strongest social motivation often comes from peers, not authority figures.

LLMs may take explanation from 90% solved to 99% solved for motivated learners. That is useful. But the deeper game changes only if someone solves motivation in a creative way. Otherwise, the technology becomes another better textbook: powerful for those already inclined to open it, inert for those who are not.

Beauty Requires Discovery, Not Just Delivery

Sanderson distinguishes between clear learning and beautiful learning. Some topics are valuable because people want to understand how they work. Others carry a particular aesthetic charge: the feeling that the universe has organized mathematical truths with surprising cleanliness.

He gives examples from higher-dimensional spheres, where the formulas for volumes unfold into unexpectedly elegant patterns, and from a collision problem where digits of pi appear where pi seems to have no business appearing. The beauty is not just that the answer is surprising; it is that the explanation makes the surprise feel necessary.

He also emphasizes the learner’s role in that beauty. As a child, he discovered that adding reciprocals of factorials seemed to approach a strange constant; only later did he learn this was connected to e. The fact felt prettier because, for a moment, it felt like his own. A teacher can show interest, but the deepest motivation often comes when the student has enough space to rediscover something personally.

Key Lessons

Why This Matters for Diffie

For Anand and Diffie, Sanderson’s “source versus relay” distinction is the sharpest takeaway. AI browser testing is becoming crowded with demos, agents, and claims about replacing QA. Diffie should not merely relay the market’s generic promise—“AI finds bugs in your frontend.” It needs to become a source of a more precise perspective: what frontend confidence means when software is changing continuously, why visual and interaction regressions escape conventional tests, and how engineering teams should think about browser-level behavior as a product quality surface.

The same lesson applies to GTM. The “algorithm” in Diffie’s world is not YouTube; it is the ICP’s attention. If frontend engineers do not click, reply, try, or retain, the answer is rarely that the market is mysterious. It is usually that the title, promise, proof, or prior trust did not give them enough reason. Replace “the market wants” with “the engineer wants” and many positioning questions get clearer.

Diffie’s best content should therefore feel like 3Blue1Brown for frontend reliability: not generic education, but visual, specific, intuition-building explanations that make a hidden problem obvious. Show why traditional E2E tests miss a class of bugs. Show a real broken flow and the exact browser behavior that exposed it. Show the mental model behind a better testing loop. The product’s credibility will compound if engineers feel, “I understand my own problem better after reading this.”

Finally, Sanderson’s team transition is a useful warning. Anand’s current challenge—ICP building, GTM motion, and outbound strategy—will create competing principles: craft the product, talk to users, publish original thinking, run experiments, follow up, and build repeatable systems. The answer is not to pick one forever. It is to write until the trade-offs are explicit, then build capacity around the bottleneck. Diffie’s long-term advantage may come from becoming the clearest source on AI-native browser testing, not the loudest relay of the AI testing trend.