Learning Rules
In the science-fiction setting of the game FTL: Faster Than Light, different species possess different abilities. Engi repair systems at 200% speed. Mantises deal 200% damage. Rockmen are immune to fire and have 150% health. Humans, by contrast, have no initial abilities; they are the most common and most unremarkable crew type. Yet they learn skills 10% faster than all other species.
I consider myself a human among humans.
Let me begin with academic learning, in contrast from skill acquisition.
In primary school, I did not know there are differences amongst ways of learning. I believed in the sufficiency of knowing the material and performing well on exams. Luckily, I was a “well-behaved student” who listened carefully in class, and my grades stayed above average. But I transferred into the international department at fourth grade: an entirely new environment. Unlike the exam-oriented system built around China’s high school and college entrance exams, it emphasized instruction in English, reading and writing, comprehension of knowledge, and the construction of a tightly interwoven conceptual network rooted in Western culture—where literature, history, and geography intersected. I was overwhelmed. My dad had anticipated this. He supervised my morning English reading and nightly review of notes, helping me stabilize my GPA within that single semester.
It was in middle school that I truly found an efficient rhythm of learning. Beginning seventh grade, second semester, I maintained first place in cumulative exam scores across five core subjects (Chinese, Math, English, Humanities, and the sciences). At the time, I was practicing badminton twelve hours per week and spending additional hours on piano and drawing. After school, I rushed to the training hall, and returned home to complete assignments only since then. Interestingly, I never felt pressed for time. Instead, my weekends still consists of generous stretches of leisure. Up until ninth grade, I never studied for STEM exams. I only take out my notebook an hour before the test and read carefully for Chinese poetry and literary knowledge and the history and geography components of humanities—and an hour was usually enough. Exam weeks were often my happiest weeks, because I did not need all the review time my teachers allocated to classmates; I could use the surplus to relax. My reasoning was that if I had fully understood and internalized the material during class, I could retrieve the chain-of-thought at any time and deploy it in the exam room. As for the one-hour-review, that can be attributed to the Ebbinghaus forgetting curve: the most recently learned knowledge has the lowest probability of being forgotten. At the time, I did not summarize any formal “learning rules.” I simply trusted my mastery of the concepts. “Paying attention in class, playing after class, reading notes before exams” has become a lifestyle.
Ninth grade marked my first attempt at self-study and exploring a method suited to myself. Because I grew up in a household with a former organic chemistry researcher, I found chemistry interesting. I earned a used Pearson IB Chemistry HL textbook from upperclassmen and read it word for word over the course of a semester, hidden beneath my desk during chemistry class. Later, I did the same with IB Physics HL, in all apparent seriousness taking thirty pages of Cornell notes covering Newtonian mechanics, circular motion, waves, and electromagnetism. I also read Mankiw’s Principles of Economics closely, along with various AP textbooks, and in tenth grade entered competitions such as the UK Chemistry Olympiad, AP Chemistry, the National Economics Challenge, and the International Economics Olympiad. By the end of ninth grade, my self-study pace was far ahead of my classes. Sometimes I ignored class content and read high school textbooks instead. My physics teacher discovered this and gave me twenty minutes to teach electromagnetic induction to my classmates.
The tension between “self-study” and “formal instruction” defined my high school learning method. On one hand, my responsible IB teachers followed a steady curriculum pace. On the other hand, I needed to study material not covered in class to compete in academic contests and AP exams. In tenth grade, I took four AP exams beyond my six IB higher-level subjects and earned 5’s on all of them. I intended to take five more in eleventh grade, but COVID canceled the exams. For two consecutive years, I participated in USAD, reading 700 pages of reference material within four months each year. For someone once hesitant about English reading, this was transformative. In tenth grade, I scored 7800 out of 10000, earning national fourth place. Slowly, I realized I had the ability and justification to acquire necessary knowledge independently. I began questioning whether the educational system I dwell in was the most efficient one. Yet the heavy IB workload left me no room to experiment with alternatives. The question lingered.
It was only in college that I encountered the term “visual learner.” During the coaching certification training of Badminton World Federation, I also learned about “auditory” and “kinesthetic” learners. I realized that my learning is purely visual. My reading comprehension speed far exceeds any other intake method. Liu Cixin once wrote in The Village Teacher, from an alien perspective, that human language transmits only 10 bits per second (more recent research suggests 39). On top of that, I believe that information is high-dimensional, but spoken language is linear. Linear transmission of high-dimensional information is inevitably slow. Speech includes filler words, pauses, hesitations. Written symbols, at least, are two-dimensional: I can process multiple keywords simultaneously, thus accelerating comprehension. This reminds me of AI’s evolution from autoregressive models to transformers. Autoregressive models predict tokens one-by-one at the end; transformers can attend to any position. The latter is computationally more efficient. My “visual learning” mode makes me long for dense text, such as books, scripts, manuals, and documentations, while disliking videos and many “interactive learning” platforms.
In college, I rediscovered the importance of mentors: they create feedback loops. The dilemma of self-study is the uncertainty—did I misunderstand something? Can I apply it? A good mentor can say “yes,” “no,” or offer subtler corrections. They can design better problem sets. If search engines and AI fail against confusions, only a human can resolve them. It took me two years to move from distrusting office hours to visiting whenever necessary. Moreover, experienced professors often possess insights beyond published textbooks. I gladly listen. Yet in early 2026, as large language models improve, I hesitate again. I can entrust many writing tasks to GPT-5o, or ask Gemini 3 Pro and GPT 5.3 Codex to write code and tests from scratch. They can generate problem sets. They can save me the ten-minute walk to office hours.
I compare academic learning to “car manufacturing”. The essence here is to assemble components (concepts) into a functioning machine. Parts without assembly (memorization without connections) produce no car. Assembly without parts (attempting understanding without concepts) as well produces no car. This analogy fuels my disdain for cram schools: why build the car twice if it can be built once? Previewing and reviewing are useful, because car manufacturing benefits from examining blueprints and performing maintenance; but the core act of construction deserves total immersion. Neuroscience suggests that memory formation depends first on depth of processing, second on intention, and not at all on duration. Profound understanding outweighs additional time. Efficiency comes from concentrated attention.
MBTI tests are not academically plausible metrics. Yet I find one of the questions once intriguing: is your personal work style closer to spontaneous bursts of energy or organized and consistent efforts? Macro-scale: organized. I plan my days upon waking. Micro-scale: bursts. Flow emerges from state-specific energy pulses.
My path in skill acquisition has been more rough. I have described how I began late in badminton, piano, drawing, and shooting, and how I did not find in one-shot a mentor or system that suits me. But since college, operating independently, I have discovered ways to accelerate my progress. Badminton is where breakthroughs came first, so I’ll base my narrative on this example.
Skill learning centers on iteration. I use versions to iOS updates (e.g., “26.x.x” or car models—Toyota Corolla, Audi A4, TT RS, RS7, R8—to describe my system update. Each iteration contains core structural upgrades that elevate overall performance.
Each iteration involves four components: reflection, engineering first-principles, deliberate practice, and testing.
I’ll start with testing and reflection. I compete in major tournaments bimonthly. After losing, I love asking opponents what I did well and poorly (and meet more players). I rarely need match footage; I remember the pain points.
From pain points, I design solutions. There always will be multiple ways to do the same thing, but not every one of them works as well. This is a trial-and-error process: I try alternatives, approximating gradient descent toward a local optimum under the current framework. Sometimes, I also observe technical footage of other players to mimick and personalize an existing solution. Performance-level can always be decomposed into problems in techniques, tactics, or physicals (e.g., backhand wrist rotation strength, high-intensity cardio from minutes 40–60, or racket-face control during drop shots). I fix them one-by-one.
Deliberate practice is the core in my skill learning cycle. Each issue demands targeted drills. This is a process with both traditionality and creativity. Since I have gone through multiple training modes and systems, and I have a large storage space in my brain, I can memorize many more ways of training and drills than my peers. Sometimes I recombine known exercises; at other occations I design new ones based on ergonomics and neuroscience. Wrist weakness can be solved with resistance bands and kettlebell rotations. Cardio requires 9 repeated sprint cycles of “50, 50, 100, 50, 50, 100, 200” (total of 7 points) that simulates a 3 sets, 21 points match structure. Correcting drop shot slicing requires practicing shadows with jumping while consciously relaxing wrist alignment to retrain neural pathways.
Deliberate practice is conscious and goal-oriented repetition and thus cognitively heavy. I visualize attention as a heat map with warm colors at focus points, such as yellowness on my wrist and grip, or greeness on potential placement spots on court (background is always blue). I also train temporal prediction, anticipating one or two shots ahead, like elite footballers anticipating seconds in advance.
Not every iteration shows immediate results. Sometimes technical gains are masked by physical deficits. The body is a system with equilibrium: though Hebbian learning strengthens synapses, neuromodulation resists imbalance. I am not better every day. But if each update ends stronger than the previous, iteration succeeds. That explains why my skill appears to outsiders to suddenly evolve at times during the year. But genotypic accumulations precede phenotypic shifts.
I am confident in my method of skill learning; this is why I kept on playing badminton. If I do not see a path to excellence, I will not pursue the mastery of that skill. That is why I put a pause on piano. I know that underlying principles transfer across skills. Mastering one learning method maps onto others. I have seen concert pianists practice measure by measure with deliberate precision in one master class. But I cannot spend enough time on multiple skills.
The current paradigm in neuroscience suggests peak cognitive condition around the age of thirty, with continued plasticity throughout life. As a twenty-years-old human being, I have a decade to iterate in badminton. If I fail to meet my highest goal (entering World Championships), I will pursue secondary goals (World Senior Championships) with continued iterations. In academics, with effective methods and hippocampal plasticity, I believe I can learn exponentially more knowledge in my lifetime than I currently know.
In my junior years, I did not get to the level many bright young players have reached; many of them already reached world top 100 at my age. I did not have their starting conditions, and at times did not even have their same order-of-magnitude of training time. Many peers began CS, Math, or STEM earlier than I did, and I admired their accomplishements in challenging seminars and theory classes. I am catching up. I believe a methodological advantage outweighs the advantage with regards to time. I can surpass them with my system. The adage goes “Mentors open the door; but cultivation depends on oneself.”
I will become my own best mentor.
Postscript: I am currently pursuing a BWF Level 1 coaching certificate. The reason is not to teach others, but to refine my own system and train myself. Supporting Brown’s badminton team is a great goal but a secondary one.