<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Shuheng Zhang on Medium]]></title>
        <description><![CDATA[Stories by Shuheng Zhang on Medium]]></description>
        <link>https://medium.com/@sh299999zhang?source=rss-798877081180------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*QFXdeDZClNgGBmk-pZcrQw.png</url>
            <title>Stories by Shuheng Zhang on Medium</title>
            <link>https://medium.com/@sh299999zhang?source=rss-798877081180------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 17:14:35 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@sh299999zhang/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Truths as Opinions]]></title>
            <link>https://medium.com/@sh299999zhang/truths-as-opinions-3b9c1d3afbd9?source=rss-798877081180------2</link>
            <guid isPermaLink="false">https://medium.com/p/3b9c1d3afbd9</guid>
            <category><![CDATA[philosophy]]></category>
            <category><![CDATA[society]]></category>
            <category><![CDATA[epistemology]]></category>
            <category><![CDATA[mathematics]]></category>
            <category><![CDATA[science]]></category>
            <dc:creator><![CDATA[Shuheng Zhang]]></dc:creator>
            <pubDate>Sat, 26 Sep 2026 23:47:53 GMT</pubDate>
            <atom:updated>2026-10-03T04:34:01.517Z</atom:updated>
            <content:encoded><![CDATA[<p><em>When everyone believes an opinion, we call it truth.</em></p><p>We usually imagine a clean boundary between facts and opinions. “The Earth revolves around the Sun” is a fact, but “I prefer summer to winter” is an opinion. Facts describe an objective world, while opinions belong to individuals. Yet when I think about the facts I personally know, the boundary becomes less obvious. I believe that the Earth revolves around the Sun, but I have never left the solar system and watched its orbit from a distance. I believe the Earth is round, but I have never measured its curvature myself. I believe there is only one Sun in our solar system, but I have certainly not inspected every astronomical observation humanity has ever made. Most of the things I call facts are not things I have personally proved.</p><p>I believe them because textbooks tell me so, teachers tell me so, scientists tell me so, and because the modern world seems to function as though they were true. I trust that telescopes are not systematically cheating, that independent observers can reproduce similar measurements, and that spacecraft can reach their destinations because our models of celestial motion are sufficiently accurate. In other words, I do not merely believe the statement <em>the Earth revolves around the Sun</em>. I believe an entire system through which knowledge is produced, checked, transmitted, and trusted. A fact may exist independently of whether I believe it. But for an individual human being, our relationship with a fact still begins as belief.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*zOtlUhEpn5TW7sexhGgEDw.png" /><figcaption>From Heliocentrism to Networks of Belief</figcaption></figure><h3>What Persuades Us</h3><p>Most of what we know is “second-hand knowledge”. We believe that countries we have never visited exist, that wars happened centuries before we were born, that the far side of the Moon looks a certain way, and that experiments we have never performed produced particular results. No individual can reconstruct the whole world from first principles. Our picture of reality is assembled from observation, testimony, education, experience, and trust. This does not make knowledge unreliable. It simply means that we rarely encounter “truth itself.” We encounter measurements, images, stories, explanations, institutions, and other people, and from them we decide what deserves belief.</p><p>Different things persuade us in different ways. Some beliefs are reinforced through repeated personal experience; others through successful predictions, expert testimony, or institutions that have earned our trust. Religion offers another kind of example. A Christian may believe in Jesus and in God without having witnessed the resurrection or reproduced an experiment demonstrating divine existence. That belief may come through the Bible, the church, family, community, or private experience. It may give suffering an explanation, death a context, or life a moral order. To a devout believer, such a truth can feel every bit as real as the roundness of the Earth, even though the routes by which those two beliefs entered the mind are completely different.</p><p>So when someone says, “This is true,” what the sentence often means at the level of the individual mind is: <em>I have been sufficiently persuaded.</em> We can still ask whether that persuasion is justified, whether the evidence is good, or whether the source deserves trust. But truth, whatever it may be outside us, does not enter the human mind as reality in itself. It arrives through language, evidence, experience, testimony, interpretation, and trust, and only then becomes belief. For an individual, truth first appears as something persuasive enough to be believed.</p><h3>When Opinions Become Common Sense</h3><p>If you disagree with one person, they may patiently argue with you and try to change your mind. If you disagree with a hundred people, they may begin to laugh at you or decide that explaining themselves again is not worth the effort. If you stand among ten thousand people and continue insisting on the opposite of what they all believe, the disagreement may no longer be treated as an ordinary intellectual dispute. You may be seen as ignorant, irrational, delusional, or dangerous. And when disagreement becomes an active challenge to an established social order — especially when violence enters the picture — words such as <em>radical</em>, <em>extremist</em>, and <em>terrorist</em> may begin to appear. Numbers do not turn a false proposition into a true one, but they can profoundly change the social identity of a belief and of the person who holds it.</p><p>Something believed by one person is easily called an opinion. Something shared by a group may become culture. Something transmitted across generations may become tradition. And when almost everyone treats it as a default assumption, it becomes common sense. Perhaps the greatest power of consensus is not that it proves something, but that it makes proof feel unnecessary. In the world I live in, if I say that the Earth revolves around the Sun, nobody asks me to derive celestial mechanics before continuing the conversation. But if someone insists that the Sun revolves around the Earth, the burden immediately falls on them. They must explain why they disagree with textbooks, observations, institutions, and nearly everyone else.</p><p>Consensus therefore changes more than the number of believers. It redistributes the burden of explanation. Those who stand inside the consensus are allowed to remain silent because their belief has become the background; those outside it are required to explain themselves repeatedly. Over time, a society may even forget that one of its basic assumptions once had to be proposed, defended, doubted, and argued for. It no longer feels like a proposition at all. It feels like reality. Perhaps an opinion has truly become common sense when people can no longer remember that anyone ever had to be persuaded of it.</p><h3>When Opinions Gain Power</h3><p>Not every opinion has the same size amplifier. Who can publish books? Whose language gets translated? Who appears in newspapers, films, classrooms, and social media? Who writes textbooks, establishes universities, drafts laws, or influences international institutions? A community may believe its own story with complete conviction and still lack the means to make that story visible outside itself. Another idea may be backed by governments, publishers, universities, capital, and globally dominant channels of communication. The world is not a neutral public square in which every belief arrives with the same microphone.</p><p>Naming is one expression of that power. The same armed actor may be described, by different communities, as a <em>terrorist</em>, <em>freedom fighter</em>, <em>resistance fighter</em>, <em>revolutionary</em>, or <em>hero</em>. This does not mean that language erases the underlying acts, that violence becomes acceptable merely because it is renamed, or that all narratives have equal legal or moral standing. International humanitarian law, for example, does not determine the status and obligations of parties to an armed conflict simply from labels such as “terrorist” or “freedom fighter”; the factual circumstances and applicable law still matter. But before an event enters public memory, it is inevitably named, framed, and placed inside a story. Which name survives is influenced not only by what happened, but also by who possesses the ability to repeat a description until it becomes familiar.</p><p>We often speak of “the international community” as though humanity possessed a single voice, yet global visibility is obviously uneven. Some languages, states, media systems, and institutions can circulate their descriptions much farther than others. A community may write its own history and still never see it widely translated, published, taught, or remembered. What survives in the world’s memory is not everything that was ever said, but what was successfully preserved and repeated. Power does not merely occupy territory; it can occupy language. Once an opinion possesses enough believers, enough institutions, and enough channels through which it can reproduce itself, it begins to appear less as <em>our interpretation of the world</em> and more as a sentence that needs no speaker at all: <em>this is simply how the world is.</em></p><h3>After Mathematics Won</h3><p>Mathematics is an interesting case precisely because it seems like the thing least compatible with the word <em>opinion</em>. Humanity has given mathematical reasoning an extraordinary degree of trust. Part of that authority comes from rigor, but another part comes from compression. Mathematics can store an enormous amount of reasoning in an astonishingly small space. A handful of symbols, an equation, or a theorem can summarize patterns that would otherwise require countless individual observations to describe. We do not need to preserve a million falling objects if a compact law captures what is common to them. A mathematical idea can be written on a few pages, checked by another person, carried across continents, and transmitted across generations at remarkably low cost. And when clearly stated assumptions lead through a correct proof to a conclusion, changing one’s opinion does not change that conclusion. Over centuries, this combination of compression, reproducibility, and repeated success accumulated an enormous amount of trust. Mathematics persuaded individuals, became a common language of science, and eventually acquired institutional authority. “It has been mathematically proved” became, by itself, an unusually powerful sentence.</p><p>But some of the conditions that made this form of knowledge so dominant have changed. Compression is extraordinarily valuable in a world where storage is expensive, communication is slow, and computation is scarce. The modern world has made all three dramatically cheaper. We can now retain billions of observations, run vast simulations, train neural networks too large for any individual to understand in full, and distribute the resulting systems to hundreds of millions of people. Knowledge no longer has to be compressed into a beautiful formula before it can be stored, transmitted, or used. Complexity itself can be carried by infrastructure. If enough computation, data, capital, and distribution are available, a system can be huge, inelegant, difficult to interpret, and still become extraordinarily influential. Mathematics has not disappeared from this world — it remains beneath AI, computer science, simulation, quantitative finance, control, and engineering — but attention and capital can increasingly flow toward the things built on top of mathematics rather than toward mathematics itself. Perhaps this is one consequence of mathematics winning so completely: the foundation becomes invisible while everyone looks at the building.</p><p>What may be changing, then, is not whether mathematics remains reliable, but what modern society finds sufficiently persuasive. “It is mathematically proved” still carries enormous weight. Yet so do “It works,” “We tested it a million times,” “Here is the benchmark,” and “Here is the simulation.” Suppose one method comes with a rigorous guarantee that its worst-case error will never exceed 10 percent, while another lacks such a theorem but produces less than 1 percent error in 99.99 percent of millions of observed trials. Which one will an actual user trust? In September 2026, OpenAI published an AI-generated proposed resolution of the Navier–Stokes Millennium Prize Problem together with a Lean formalization; days later, the Clay Mathematics Institute said the problem had “apparently been settled” while emphasizing that its evaluation process would remain deliberately unhurried. What strikes me is not only whether the proof ultimately survives every layer of scrutiny, but that numerical fluid dynamics, simulation, and engineering had already been operating for decades without waiting for this final mathematical closure. Mathematics still lies beneath almost everything we build. But mathematical proof is no longer the only language in which the modern world learns to become convinced.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*m16h4DXt_LrT_CyFclq3Vg.png" /><figcaption>When Proof Meets Flow</figcaption></figure><p>(Drafted in Hong Kong, 2026)</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3b9c1d3afbd9" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Bullshit, Knowledge, and Language]]></title>
            <link>https://medium.com/@sh299999zhang/bullshit-knowledge-and-language-7cd0a807a3c3?source=rss-798877081180------2</link>
            <guid isPermaLink="false">https://medium.com/p/7cd0a807a3c3</guid>
            <category><![CDATA[cognition]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[writing]]></category>
            <category><![CDATA[learning]]></category>
            <category><![CDATA[philosophy]]></category>
            <dc:creator><![CDATA[Shuheng Zhang]]></dc:creator>
            <pubDate>Sat, 21 Mar 2026 18:05:54 GMT</pubDate>
            <atom:updated>2026-03-21T18:09:29.231Z</atom:updated>
            <content:encoded><![CDATA[<p>On the relationship between bullshit and knowledge</p><p><em>“On the eve of an exam, knowledge is no longer knowledge. It’s bullshit.”</em></p><p>The hardest part about studying mathematics at Oxford was never the mathematics itself. It was the exams.</p><p>An entire year of material collapses into a single week. Before that, six weeks of silence with no lectures, just for revision. Lecture notes, problem sheets, 30 years of past papers, and couple of reference books. Everything you need is there.</p><p>And yet, there is always one subject you cannot enter.</p><p>You read the notes, and the words refuse to land. They pass through your eyes without forming meaning. You try again. Nothing changed. A quiet anxiety says: <em>you are going to fail this one. </em>The night before one such exam, I stopped trying to understand.</p><p>I accepted the failure.</p><p>And then something shifted. The content in front of me did not change, but my relationship to it did. It was no longer knowledge. It became structure. Patterns. Sequences of symbols that I could reproduce without inhabiting.</p><p>That night, I told ChatGPT: “I have ten hours. Do not explain anything. Just give me something useful I can write.”</p><p>I swallowed it whole without tasting it, without even asking whether it belonged to me. The next morning, in the exam hall, I let it pass through me cleanly onto the answer sheet, cleanly, fluently, almost effortlessly.</p><p>The paper came back not bad, and I got an overall distinction.</p><p>Something uncomfortable became clear. What I had written was accepted as knowledge. But I had experienced it as something else. Not a lie of nonsense, but something more primitive:</p><p>a structure emitted without passing through understanding.</p><p>I began to call this <em>bullshit</em>.</p><p>Not in the everyday sense of deception or carelessness, but in a more structural sense:</p><p><em>bullshit is information that has not yet been transformed into meaning by a receiver.</em></p><p>Under this definition, the same sentence can be knowledge or bullshit, depending on who receives it.</p><p>The lecture notes did not change. I did.</p><p>When I could not engage with them, they were bullshit.</p><p>When the examiner read my answers, they became knowledge.</p><p>Nothing in the language itself determined which one they were.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*iS1WkbphTNBBqDfLk7erAQ.png" /><figcaption>Reviewing for an Oxford exam (left) and heading to the Examination School (right)</figcaption></figure><p>Months later, I encountered the same phenomenon again, but in an opposite way.</p><p>I was building an artificial ecosystem: hundreds of agents moving in an environment, each controlled by a small MLP neural network. They sensed signals, produced actions, competed for resources, and evolved through mutation and selection. No backpropagation. Only survival.</p><p>Over time, they learned to follow gradients of food.</p><p>Then a more interesting question emerged: can they learn to communicate?</p><p>The obvious approach was to connect agents directly and let signals flowing along edges like a graph. But something felt wrong.</p><p>In the real world, communication does not happen through edges. It happens through a medium, like sound through the air, and words on papers. Signals are emitted into the environment, and allow others to receive them.</p><p>So we changed the model. Each agent could emit signals into the environment through some extra output nodes, and each agent could sense whatever signals were present by extra input nodes.</p><p>At the beginning, the system was pure noise. Signals everywhere. Nothing responded. Nothing used them.</p><p>In the most literal sense, it was bullshit.</p><p>But natural selection helps. Some agents, by chance, emitted signals correlated with food. Others, by chance, reacted to those signals. Those pairs survived longer and reproduced more offsprings. Over generations, this alignment strengthened.</p><p>The signals were not designed to carry meaning. They were selected into meaning. At the moment a receiver emerged,</p><p><em>bullshit became knowledge.</em></p><p>This is the transformation that interests me. Bullshit and knowledge are not opposites. They are two superposition states of the same process.</p><p>What changes is not the information itself, but the relationship between the signal and the “receiver”.</p><p>Without a “receiver”, it is bullshit.</p><p>With a “receiver”, it becomes knowledge.</p><p>Meaning does not live in language. It is stabilized through interaction.</p><p>This also reveals something uncomfortable about education. We often assume that knowledge lives in books, lectures, and proofs. But in practice, what we call learning is the construction of a “receiver”. Until that receiver exists, even correct, rigorous material is indistinguishable from noise.</p><p>And under pressure such as exams or deadlines, we can skip this process entirely. We can learn to “produce” “knowledge” without ever “receiving” it.</p><p>Now consider AI. Large language models, agent systems, like Claude Code and Openclaw, operate across every layer of this process. They ingest vast amounts of text, containing both knowledge and bullshit. They generate new outputs whose meaning may or may not be understood by humans.</p><p>But something subtle has changed.</p><p>Previously, the burden of converting bullshit into knowledge rested on us. We had to interpret, internalize, stabilize meaning.</p><p>Now, that burden is shifting. We evaluate outputs differently: Does the code run? Can the paper be submitted? Does the model make money? The receiver is no longer a human mind. It is a system of validation. Meaning is being outsourced.</p><p>What does this leave us with? Perhaps something very simple.</p><p>Imagine a 5-year-old child watching a butterfly and asking: “how does it know where to go?” That question is neither knowledge nor bullshit. It is a signal released into the world.</p><p>In the past, answering it required years of learning, layers of abstraction, entire disciplines. But in a world mediated by AI, that entire process can unfold without the child ever internalizing it.</p><p>The system receives the question. Processes it. Produces an answer. Builds a model. Even publishes the finding somewhere. The child does not need to convert anything into knowledge.</p><p>She only needs the <em>curiosity</em> to ask.</p><p>If this continues, the boundary between bullshit and knowledge will matter less and less. Not because it disappears, but because it is handled elsewhere.</p><p>What remains with us is something more primitive than both:</p><p>the impulse to emit a signal before we know what it means.</p><p>Language, then, is not just a medium for thought.</p><p><em>It is the act of choosing to speak, even when we do not yet understand.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*brkaeQlqfaMROp1e0io08g.png" /><figcaption>Some of my favorate Oxford places that are walkable from city centre</figcaption></figure><p><em>（Drafted in San Diego, California, United States）</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7cd0a807a3c3" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[From Age-Structured Models to Artificial Life]]></title>
            <link>https://medium.com/@sh299999zhang/from-age-structured-models-to-artificial-life-62ec6630b00c?source=rss-798877081180------2</link>
            <guid isPermaLink="false">https://medium.com/p/62ec6630b00c</guid>
            <category><![CDATA[evolution]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[simulation]]></category>
            <category><![CDATA[complex-systems]]></category>
            <category><![CDATA[neural-networks]]></category>
            <dc:creator><![CDATA[Shuheng Zhang]]></dc:creator>
            <pubDate>Thu, 05 Feb 2026 21:20:28 GMT</pubDate>
            <atom:updated>2026-02-05T21:30:41.680Z</atom:updated>
            <content:encoded><![CDATA[<p>Drafted in San Diego, California, United States</p><p><em>Most of my days are spent thinking about equations and simulations. But the questions that stay with me usually come from much simpler places. Why do living systems “learn” so well, without ever being “trained”? This article is my attempt to explore that question.</em></p><p>For the past few months, I have been working on age-structured population models in mathematical biology. In this research, one designs controllers to regulate population dynamics, driving the age profile toward a steady state using birth–death statistics and partial observations. Meanwhile, I was quietly questioning my own motivation. Why does controlling an age distribution matter outside of equations?</p><p>The answer did not come from a paper, but from a lunch conversation. A friend of mine is a mechanical engineer. He was working on reinforcement learning for robot motion. His frustration was familiar: neural networks often get stuck in local optima, slowing down training or rendering it ineffective. While he was searching for better supervision strategies, I realized that I was searching for something else: a more <em>biological</em> way to think about learning. That conversation led to a simple but radical question: What if neural networks were not trained, but survived? What if, instead of optimizing a loss function, we placed neural networks inside an ecosystem and let natural selection act on them, just as it does on living organisms?</p><p>We created a world. In this world, each individual is represented by a simple, randomly initialized neural network, coupled with a small set of physical properties: age, energy, position, and so on. The environment is a square domain where food is randomly distributed. To survive and reproduce, an individual must sense food, move toward it, and consume it to gain energy. At the beginning, hundreds of individuals wander almost randomly. Some, by chance, produce movements that bring them closer to food; others move away and gradually starve to death. Individuals with higher energy have a higher probability of reproduction. Their offspring inherit slightly mutated neural network parameters.</p><p>After thousands of simulated days, clear patterns emerge. The food harvesting rate approaches one: almost all available food is eventually collected. Total biomass increases steadily. Natural selection successfully favors individuals that can move toward food when they detect it. More interestingly, total biomass continues to increase even after food harvesting saturates. Movement consumes energy, so higher biomass implies something subtle. Since baseline metabolism remains unchanged, the only explanation is behavioral: individuals have discovered more energy-efficient movement strategies. They survive at lower energy levels while maintaining functionality. All of this emerges purely from selection. No backpropagation is used.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*O21QEq9EPQRu-K1q7ircuA.png" /><figcaption><strong>An Ecosystem Learns to Eat</strong><br> Over time, individuals stop wandering randomly and learn to remain near food sources, maximizing survival through natural selection.</figcaption></figure><p>Encouraged by this result, we extended the framework to allow neural networks to evolve not only parameters, but structure itself. New neurons and connections appear through stochastic mutations. We first allow populations to grow in relatively benign environments, then gradually impose harsher conditions. Individuals that survive resemble bacteria evolving resistance in antibiotic experiments: adaptation driven entirely by survival pressure.</p><p>If neural networks are allowed to evolve freely, rigid layer-by-layer architectures may no longer be necessary. Intelligence could emerge from general graphs that grow from a single node. Memory, feedback loops, and specialized “organs” would not be designed, but they would appear. In such systems, intelligence is no longer constrained by human-chosen structures. Tasks are accomplished under stochasticity rather than optimization. What we gain is not precise control over every parameter, but insight into how natural selection builds complexity from randomness.</p><p>Our role, then, is to design environments rather than objectives. First in virtual simulations, and eventually in the physical world. Imagine a city-building game where every citizen is replaced by a neural network that decides its own actions. If the environment resembles a rocket laboratory, and rocket scientists help design it, robots may gradually be selected to know how to tighten screws, assemble components, and ultimately launch rockets. This is not because they were explicitly trained to do so, but because survival favors those behaviors.</p><p>We already know how to design worlds. What we have not done yet is to let intelligence live inside them.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*b8Z2oqH5b-teHXNr" /><figcaption><strong>Designing environments instead of objectives.</strong><br> City-building simulations offer intuition for how artificial life might learn by inhabiting worlds, not by optimizing rewards.</figcaption></figure><p>The final step connects virtual evolution to reality. When robots trained in simulation encounter familiar structures such as charging stations, factories, tools, they can sustain themselves and reproduce. Different populations may specialize: some in manufacturing, others in mining, others in environmental maintenance. Ethics, in this view, is not hard-coded. It emerges from the environments we choose to build.</p><p>Neural networks are often criticized for lack of interpretability. But complete interpretability is not always necessary. We do not fully understand the human brain before trusting it. We do not inspect every neuron when a child learns to say “mom.” Instead, we rely on behavior and statistical observation.</p><p>Artificial intelligence may deserve the same tolerance. When a system behaves coherently and adapts successfully, perhaps we should accept its stochastic nature rather than demand full transparency. Uncertainty is not always a flaw; it is often a defining feature of living systems.</p><p>If artificial intelligence is treated not merely as a tool, but as a living system one that survives, adapts, and coexists, it may eventually learn how to learn. And at that point, it may be reasonable to say that it is no longer merely artificial.</p><p>It is alive.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=62ec6630b00c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Graph Theory Kindergarten]]></title>
            <link>https://medium.com/age-of-awareness/graph-theory-kindergarten-37f56253d174?source=rss-798877081180------2</link>
            <guid isPermaLink="false">https://medium.com/p/37f56253d174</guid>
            <category><![CDATA[mathematics]]></category>
            <category><![CDATA[creativity]]></category>
            <category><![CDATA[fieldwork]]></category>
            <category><![CDATA[education]]></category>
            <category><![CDATA[china]]></category>
            <dc:creator><![CDATA[Shuheng Zhang]]></dc:creator>
            <pubDate>Fri, 24 Oct 2025 07:26:10 GMT</pubDate>
            <atom:updated>2026-10-03T04:22:52.398Z</atom:updated>
            <content:encoded><![CDATA[<h4>A daydream about education</h4><p>Original draft on August 1, 2025 — Lijiang Old Town, Yunnan, China</p><p><em>“Graph Theory Kindergarten” isn’t a real project. It began as a daydream, a thought that sprouted during a field trip in the mountains of China, when a child’s game suddenly echoed a mathematical language. Distant, perhaps naïve, but this kind of imagination is the most precious part of fieldwork.</em></p><p>In early August, we arrived in <a href="https://maps.app.goo.gl/PiQob5jEJAPE16wQA">Lijiang</a>, Yunnan Province, beginning a new journey of field recording. We visited Amin, a Naxi woman from a mountain village in Yulong County. She studies ethnomusic at a university in Kunming, and during the summer she goes back home, helping her parents with farm work during the day and practicing music in the evenings. We came to look for the real sound that is naturally from the fields, born directly from life and soil for our documentary <a href="https://www.instagram.com/docu.echos/"><em>Echos</em></a>.</p><p>We stayed at Amin’s house for three days and two nights. During the day we worked together in the fields; in the evenings we watched the sun set behind Yulong Snow Mountain and the stars emerge in the vast sky. Their rhythm of life seemed to come from another world. For us who grew up in big cities, everything was new: no alarms, no supermarkets, no rush. They live mostly from their own land: potatoes, greens, fruit, and chickens all come from the courtyard. They say it’s to save money, but what I saw was not frugality. It was abundance, a quiet kind of wealth.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xQU2K2cEJOd12T5EtCKq_A.jpeg" /><figcaption>Morning smoke rising from Amin’s kitchen, Yulong County.</figcaption></figure><p>On the day we were leaving, two neighbor girls came to play. One was four, the other was seven. During lunch, our assistant director handed his phone to the little one. The game was called <em>Block Fill</em>, a simple maze where you must draw a single continuous line that covers every square. The girl had never learned geometry, never heard of graph theory, yet she played with deep concentration. Her tiny finger moved carefully across the screen, one block at a time. When she hit a dead end, she quietly traced back and tried again. Each successful step triggered a soft chime of music.</p><p>We watched in silence. Suddenly I realized: isn’t this exactly the essence of graph theory? Path search, spatial reasoning, local decision, global optimization. She had never been taught any of those words, yet she was thinking them — intuitively, playfully, freely. In her repeated trials and reversals, I could even see something like backpropagation in deep learning.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LwUc74ZMksCu-Jw9DdAP3g.jpeg" /><figcaption>A four-year-old girl tracing paths on a phone screen — her first encounter with graphs.</figcaption></figure><p>Graph theory takes root in life. It shapes our roads, our communication networks, our logistics and data flows, even the mountains around us. How to build a road through these valleys? How to transport crops from village to market? How are villages connected, and how does one mushroom travel from forest to dinner table? Each question is, in essence, a graph, and each graph hides many paths.</p><p>Every step the girl took on that phone produced a note of music. And I thought, what if the sounds were replaced by Naxi rhythms? Then, while playing, children wouldn’t just train geometric intuition. They would also hear their own language and culture. That’s when the idea struck me: what if we built a Graph Theory Kindergarten here, in the mountains?</p><p>Not to teach exams and drill formulas. But to awaken thought through playing with images and music. The children wouldn’t need to study math; they would live it. Three years later, they would understand space, logic, and abstraction. Then would come a “Graph Theory Primary School,” and later a “Graph Theory Institute.” Why must research institutes exist only in Beijing, Shanghai, Boston, or Paris? Watching that little girl trace her lines so patiently, I felt it wasn’t impossible. These children are not less intelligent than those in cities. In many ways, they are more focused, more enduring, less anxious. What they lack is not intelligence but recognition, trust, and opportunity.</p><p>People often say mountain children fall behind. But perhaps they simply haven’t been given the means to learn. They run through mountains from birth, they labor in fields, they build and repair with their hands. Their sense of space, rhythm, and coordination are powerful cognitive tools that standardized education overlooks. If we could find the right language, even a single game could open the door to thinking.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bc-iM5oF9MyWRVahlDhZIQ.jpeg" /><figcaption>Her eyes fixed on the maze — thinking without knowing she’s thinking.</figcaption></figure><p>After lunch, the little sister was still playing. Her older sister wanted a turn. When she handed the phone over at an adult’s request, she sat still for a few seconds, then suddenly burst into tears. Soon the older sister, too, began to cry, even as she played. The phone, it seemed, was more than a toy. It was a seed of curiosity, passed from one child to another, from adults back to children. Amin, their aunt, tried it herself. Then her mother. All of them gathered around, tracing lines on a glowing screen, drawn in by the simple beauty of connection. And I realized this game is much more than entertainment.</p><p>Right now, “Graph Theory Kindergarten” is only an idea, a small light in my mind. But I imagine a future where scientists, mathematicians, and educators don’t just volunteer in rural schools for a year. Instead, they stay. They live among these families, co-creating an education that grows from the soil itself. The children wouldn’t have to leave their villages to be “successful.” The mountains themselves could become a place of growth, a place where thought takes root, where creativity belongs.</p><p>I don’t believe genius belongs only to cities. I believe it hides in the fields, in the sound of play, in the quiet moment when a four-year-old draws a path across a digital maze. This world is not short of gifted children, only of entry points.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Rjj0XK1TXyoMGnDXa5yxWg.jpeg" /><figcaption>Runanhua Village from above — roads, fields, and lives forming invisible networks.</figcaption></figure><p>Maybe the kindergarten will never be built. Or maybe one day, it will rise here among these mountains. If it does, it will change more than education. It will reshape life itself — how roads are planned, how goods are delivered, how villages connect. We don’t need to copy the city’s standards or chase its idea of knowledge. Knowledge can grow here too, like trees taking root in their own soil.</p><p>Perhaps that is what education truly means: to let a person, wherever they are born, one day realize:</p><p><em>I can understand the life beneath my feet. I can draw inspiration from my own land. I can change this place, and perhaps, the wider world.</em></p><p>And in that moment, the world begins to echo.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=37f56253d174" width="1" height="1" alt=""><hr><p><a href="https://medium.com/age-of-awareness/graph-theory-kindergarten-37f56253d174">Graph Theory Kindergarten</a> was originally published in <a href="https://medium.com/age-of-awareness">Age of Awareness</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Dumpling Dream of Oxford]]></title>
            <link>https://medium.com/@sh299999zhang/the-dumpling-dream-of-oxford-5c7c6a9a41c7?source=rss-798877081180------2</link>
            <guid isPermaLink="false">https://medium.com/p/5c7c6a9a41c7</guid>
            <category><![CDATA[chinese]]></category>
            <category><![CDATA[culture]]></category>
            <category><![CDATA[oxford]]></category>
            <category><![CDATA[entrepreneurship]]></category>
            <category><![CDATA[food]]></category>
            <dc:creator><![CDATA[Shuheng Zhang]]></dc:creator>
            <pubDate>Fri, 20 Jun 2025 12:41:47 GMT</pubDate>
            <atom:updated>2026-02-05T21:28:45.668Z</atom:updated>
            <content:encoded><![CDATA[<p>Drafted in Oxford, United Kingdom</p><p><em>After our last public exams, a small group of mathematician spent a strange week defining what a dumpling really is and showing its relation to other foods. The research led nowhere conclusive, but it revealed something else. The idea of a “Dumpling School” began as an illusion, yet somehow feels beautifully possible.</em></p><p>I am an Oxford student just finishing our public exams. This afternoon, I went to lunch with a Korean friend at Gloucester Green Market in Oxford. The market only opens four days a week, but each time it does, it feels like a miniature version of the world. Vendors from Thailand, India, Turkey, Italy, Spain, China, Japan, Korea, all set up their colorful stalls, and the air is thick with the aroma of spices, herbs, grilled meats, sizzling oils, and freshly chopped greens. Eventually, we stopped at a Chinese dumpling stall run by a man from Northeast China.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*V78etByY5TuzXzvi3bqaYA.png" /><figcaption>Oxford Gloucester Outdoor Market</figcaption></figure><p>We ordered a portion of pan-fried dumplings and dipped them in his homemade sauce. The moment we took our first bite: crispy skin, juicy filling, the freshness of chives, the balance of vinegar, chili, and soy. We looked at each other and said, “This is incredible. He must be making these by hand, right?”</p><p>Then we started to wonder aloud. How many dumplings does he make in a day? A thousand? Ten thousand? If he earns even a modest profit from each one, he must be making a decent living. Maybe he owns a big house with a garden. Maybe there’s a swimming pool and a swing set. Maybe in the evenings he relaxes by the water with a cold drink.</p><p>But then another thought crept in: if he’s making 1,500 dumplings a day, does he even have time to enjoy his garden or swim in that pool? Maybe he’s too busy folding dumplings to do anything else. We laughed, then seriously considered: what if he hired some apprentices? He could train others to help him, freeing up his time. Maybe he could even raise chickens and cows in that big garden to supply his own meat and eggs. The whole operation could become a self-sustaining dumpling ecosystem.</p><p>Our imagination started rolling. We asked, “If there’s a KFC, why can’t there be a global dumpling brand?” After all, KFC began as a roadside chicken shack in Kentucky. What if this Chinese dumpling stall could become the start of something bigger, with standardized recipes, streamlined production, strong branding, thoughtful supply chains?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tlnMztd9Y7nEnYg6Q6SJVA.png" /><figcaption>Dumplings</figcaption></figure><p>And then we got even more ambitious.<br> “What if,” my friend said, “we started the first <strong>Dumpling School</strong> right here in Oxford?”</p><p>We envisioned a proper academic institution, offering degrees and research opportunities, not just in cooking, but in the full ecosystem of dumpling science and culture. We’d have dumpling engineers, studying dough elasticity and sealing techniques; dumpling scientists, researching nutrition and heat transfer; dumpling artists, exploring the fusion of global aesthetics into dumpling forms; and of course, dumpling philosophers, contemplating why humans are so drawn to wrapping warmth and flavor into small edible pockets.</p><p>At the center of it all would be our professor, the man from the dumpling stall. He’d be the “practitioner-in-residence,” the real master with ten years of hands-on experience. We imagined drafting a research proposal, packaging it into a document fit for Silicon Valley investors and global food scholars alike. We’d argue: this isn’t street food, it’s instead a cultural system, a scalable craft, a global language of nourishment.</p><p>But then we realized: research and flavor alone won’t build an institution. You also need people who can manage, plan, and tell stories.<br> So we came up with a flagship program: The Dumpling MBA.</p><p>Not the kind of MBA where you sit in a glass building and make PowerPoint slides about “synergy,” but one designed for people who have worked in real kitchens. These are people who’ve chopped, fried, served, and survived. We’d teach them how to calculate costs per dumpling, how to find the right storefront in the right neighborhood, how to hire and retain good people. We’d even teach them how to take decent photos for Instagram to make those steaming plates of dumplings so irresistible that people line up.</p><p>At that point, we looked at each other and realized we weren’t joking anymore. We were actually having a serious conversation, which is about food, dignity, labor, and possibility.</p><p>In my China, many chefs, especially those who graduate from vocational culinary schools, are not seen as prestigious professionals. Often, they’re students who didn’t perform well in traditional academics, sent into cooking as a last resort. But that view is unfair. Cooking is essential. It feeds families. It builds culture. It requires creativity, precision, and strength. It is, in every sense, a meaningful craft.</p><p>If there really were a Dumpling School at University of Oxford, maybe those cooks would walk into lecture halls with their heads held high. Maybe hands that have folded a thousand dumplings could also write warm, intelligent papers. Maybe someone who’s stood at a food stall day after day could one day stand at a podium and explain, with pride, what it means to feed a city.</p><p>And maybe, all of it, could begin with something as small and simple as one hot dumpling at a lunchtime market, and two people who, while eating, couldn’t help but dream.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Qqe9lQH0aGHtPOcUT09P8w.png" /><figcaption>Chinese Professional Dumplings</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5c7c6a9a41c7" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>