We help learners achieve much higher levels of ability in STEM, in less time, with more purpose and enjoyment.
Our mission
Making a world of great problem solvers
Our approach
At Brilliant, we design and test learning that’s interactive, adaptive, and fun — at scale.
We’re empiricists. We measure everything, and at least one human on the team still reads every piece of qualitative feedback we get.
For a learning product, the proof is in the pudding. It’s straightforward to measure things like how many users would (and do) practice every day, and whether people successfully complete problems of increasing difficulty over time.
With well over 1 million problems solved per day, we’re rapidly increasing the pace at which we’re able to test our pedagogical ideas.
Getting the learning experience just right is what’s driving this exponential growth.
What’s the ideal way to teach the concept of a variable to a beginner? At Brilliant, this is the kind of question that everyone (math PhDs, International Math Olympiad medalists, HCI designers, educators, artists, and engineers) obsesses over.
We sell directly to learners, rather than distributing to schools and workplaces that force people to use the product. We grow on the strength of being so good that you’ll voluntarily learn each day. This is very hard to do for math. But we’re demonstrating that it’s possible.
Our learners
Demographically, we teach a broad range of users. We serve students seeking to excel in STEM, professionals learning new skills or refreshing dormant ones, and lifelong learners staying sharp.
As an ethos, we teach people who take pride in having a well-trained mind.
Brilliant’s recent ad campaign highlights the power of a well-trained mind.
While learning on Brilliant should directly translate to real-world results (e.g., being able to solve similar problems on a test or at work), we see this as an outcome, not as the core purpose of learning.
Problem solving, like play, is a natural instinct. Humans love to figure things out. We aim to feed this intrinsic motivation by making learning fun and creating a lifelong habit of self-challenge. This can happen at any age or stage of life.
We believe that realizing the potential of your mind is in itself a satisfying pursuit.
What we teach
Brilliant is primarily focused on mathematical and quantitative problem solving, and more recently computer science.
In math, by the end of 2025 we will cover all core concepts of Algebra.
In 2026, we will build out our coverage of Arithmetic, Geometry, Probability, Calculus, and beyond.
In CS, by the end of 2026 we will cover all of foundational algorithm design and data structures, from two perspectives:
The “programming-first” perspective (writing and running programs)
The “programming-last” perspective (structuring logic to ensure correctness and efficiency)
This will bridge complete beginners from their first program all the way through the end of college-level introductory computer science courses.
Our method
We continually test and optimize learning delivery in these 5 areas:
Our lessons prioritize active learning.
Each lesson is typically focused on a single concept. The lesson has a mix of blocked problem solving (i.e., reps on the same or similar type of problem) and direct instruction in the solutions and in teaching panes.
When first introducing a concept, we build intuition with visual explanations, hands-on manipulation, and concrete computation. We introduce the easiest version of the idea first — so we minimize cognitive load with the simplest version of computation. Our interactives give you instant, custom feedback based on how you correctly or incorrectly solved the problem. They also actually fully simulate what’s happening: Our code environments actually run your code, our circuit interactives actually model a real circuit, etc., rather than relying on canned outcomes.
As concepts in later lessons build on earlier lessons, we increasingly challenge learners to combine multiple ideas to solve problems, or perform more difficult computations.
We currently do not employ the strategy of teaching you the procedure before asking any questions. Instead, we pretest on the material (i.e., provoke the learner to try the question without instruction) before teaching it mid-lesson. Based on our testing to date, we believe this is the optimal approach.
Every lesson has associated practice sets. These practice sets are designed to feel like low-stakes quizzes. The frequency, timing, and composition of practice sets are an area of active experimentation, to maximize effective retrieval and automaticity.
In practice sets, the scaffolding falls away — you’re being tested on your independent ability to answer the questions, so there are no more visual aids or hints.
In some courses, we’re testing (with positive results) changing problem framing. So, for example, in our introductory Statistics course the lesson asks you to find the mean of a quantity of cupcakes sold. Then the associated practice asks you to find the mean of the sale price of a drink.
For constructing practice sets, we’re continually experimenting with ways to predict the optimal next problem X to ask, so that you are adequately prepared to answer a future question Y. This is a relatively more generalized testing umbrella into which we organize efforts into spaced repetition (especially in weak areas), mixing practice problems from different concepts (so that the learner must identify what approach to use to solve each problem, rather than just applying the same procedure to every problem), determining level of effective automaticity (fast solving speed with no mistakes), and optimal length and difficulty per set.
Review sets that combine problems from preceding lessons will roll out this Fall for all users, on a course-by-course basis (we human-review everything, which is why this has a gradual rollout).
In 2026, we’ll also experiment with timed testing to more closely mimic higher-stakes situations.
To learn math well requires many, many reps. The precise amount required varies among individuals, but everyone needs practice. We design a variety of mechanisms to make it fun to practice, and satisfying to progress to more challenging sets.
Our gamification layer provides scaffolding for motivation and habit formation. This provides another tool to help learners stick with learning even in the absence of teachers who help foster motivation, tutors or coaches who enforce a disciplined daily practice schedule, or parents who provide corrections and rewards.
On Brilliant, the content itself is designed to feel fun — from the pedagogical design of the interactives and problem progression, to the types of sounds, feedback, and haptics we provide as you interact.
Most of the increase in usage and learner retention has come from excellent content delivery, but gamification helps. We avoid packing too many game incentives into the product to keep the focus on the learning content, but aim to execute well on a few core habit formation loops. These include Streaks and Leagues.
Learners progress through levels of progressively more difficult problem solving in a topic. The leveling system intentionally displays a more reductive view of concept connections and prerequisites than the underlying reality. It is focused on providing learners with clear and satisfying milestones to work toward.
Progress is also gamified — including in how we structure XP and pass/fail feedback and routing. Currently, we’re working on testing various mechanics to extend your learning session, i.e., motivate learners to continue with the next appropriate practice, review, or lesson.
Overall, approaches to building both intrinsic and extrinsic motivation are an area of active investigation and testing.
Brilliant is available on four platforms: desktop web, mobile web, iOS, and Android.
We sweat the details of removing as much from the interface as we can (including killing features that test only incrementally better), and making every UI element, interaction, and feedback obvious without needing a tutorial.
Every 30 ms we shave off of our loading and feedback times makes a difference in how much people learn, and to our personal satisfaction with the performance and craft in our engineering. For mature features, we routinely revisit how close we are to the limits of performance, and will rewrite everything if necessary to get closer.
It’s a core part of our business model today to allow everyone to learn something every day, for free. As we continue to build more features into our learning product, our aim is to make the free tier more generous, with various subscription tiers offering additional learning features for individuals and homeschools.
In Fall 2025, we will begin rolling out free learning videos for concepts covered in our courses, starting with our foundational math and CS courses.
Alongside adaptive practice delivery, ‘true’ personalization at the level of a 1:1 human tutor is an area we are actively user testing.
AI models bring us much closer to being able to have a multi-modal conversation in real-time with the material, but there are several important problems to solve:
User knowledge modeling. Teaching well requires having an understanding of what the learner does and doesn’t know. We’re combining techniques from intelligent tutoring systems, classic ML recommender systems, and natural language conversations to identify exactly where a learner’s misconception lies.
On-the-fly visual and interactive generation. Learners get lost in a wall of text. They want a visual or a manipulable interactive, problems that exactly match the question they have, and the ability to practice on near-neighbor problems just like it. Our content has been built over many years to enable precisely this functionality.
Interface. What should it look like, and more importantly feel like, to have a conversation with a human tutor? We know what it shouldn’t look like: the tutor giving you three paragraphs of text to read, just giving you the answer, or offering hints so easily that it becomes a crutch. We’re charging toward the future of what this should look like. If you’re a UI/HCI designer and this sounds exciting to you, please reach out!
Math isn’t merely the least popular school subject — for many, the mention of ‘math’ triggers fear and anxiety with a vehemence that puts it in a class of its own. Read more.
Math has a hierarchical subject structure, where concepts build on each other. In geography or history, topics can be learned more independently. But in math, there is a long dependency chain of prerequisites to master before learners can successfully learn the next idea.
Mastery and progression to higher levels of math requires cognitive struggle. Much of learning research essentially boils down to: Learning methods that impose struggle are more effective. But when you’re struggling, it feels like you’re learning less. So it’s really hard to motivate people to struggle. And even for those who choose to try, it’s within the modern context of a world that eats away at long attention spans, slow, methodical thinking, and intense bursts of concentration on something difficult.
So, we are battling: A world full of people who hate math, who mistakenly believe that the most effective ways of learning are not effective for them, and who are addicted to algorithmic media designed to rot your brain.
We are determined to prevail against these long odds. If this also sounds like you — please check out our Careers page and reach out!
Careers
Come work alongside a team of diverse talents – multiple IMO medalists, ex-Editor-in-Chief at The Onion, Cannes Lion winner, knitwear designer for Marvel, an Amazon top 50 book of the year writer, and PhDs and dropouts from MIT/Caltech/Stanford/Harvard, to name a few.
