Three substrates now share that work — biological brains, computational chips, and the markets that price both. Their relationship is becoming the central dynamic of the modern world. 三种基底分担此功——生物大脑、计算芯片、与定价二者的市场。三者的关系正在成为现代世界的中心动力。
This site analyses cognition as civilizational infrastructure: how brains and chips compare, how scientific and commercial markets price intelligence differently, why GPUs became strategic, and where the structural inequalities sit. The framing is anti-elitist — talent is treated as a system property, not as biological destiny — and explicit about where speculation begins. 本站将认知视为文明基础设施加以分析：脑与芯的对比、科学与商业市场对智能的不同定价、GPU 为何成为战略商品、结构性不平等之所在。框架反精英主义——将'才能'视为系统属性，而非生物命定——并对推测之始处保持透明。
A neuron is not a transistor. But at the abstract level — receive weighted inputs, compute, send a weighted output, update the weights — the resemblance is deliberate. Modern neural networks were called "neural" because their architects were copying the abstract job description, not the substrate.神经元不是晶体管。但在抽象层面——接收加权输入、计算、发出加权输出、更新权重——其相似性是有意而为之。现代神经网络被称为"神经"，因为其架构师所抄录的是这份抽象的工作说明书，而非其基底。
| Axis维度 | Biological brain生物脑 | Modern accelerator现代加速器 |
|---|
From bacteria to AI, intelligence has shipped in seven distinct architectures. Each adds one more capability to the previous; none replaces it. Modern minds are still running every layer, all the time.自细菌至 AI，智能以七种不同架构发布。每一种都为前者增加一项能力；没有任何一种取代前者。现代心智在每一时刻都同时运行所有层。
"Mathematical talent" is a useful shorthand for a configuration that emerges from training, environment, and specialisation — not a fixed substance possessed by a small minority. Six structural points are worth holding in view simultaneously."数学才能"是一个简称，用来指一种由训练、环境与专门化所涌现的配置——并非少数人天生具有的固定物质。以下六条结构性观点值得同时保持在视野中。
Why does a frontier-AI engineer earn ten times what a Fields Medalist makes? Not because their cognition is ten times more valuable in any deep sense. Because of time horizons, infrastructure leverage, network effects, and the simple fact that prestige and cash are different currencies. None of these is a moral verdict.为何一名前沿 AI 工程师的收入是菲尔兹奖得主的十倍？并非其认知在某种深层意义上贵十倍。原因在于时间尺度、基础设施杠杆、网络效应，以及一个朴素事实——声望与现金是两种货币。其中没有一项是道德裁决。
The history is short and the implications are large. A class of arithmetic — many small parallel multiply-adds — went from rendering video games in 2005, to underwriting most of contemporary scientific computing by 2015, to running the world's largest cognitive infrastructure by 2025. Five steps in twenty years.这段历史短而其含义大。一类算术——大量并行小乘加——从 2005 年的渲染游戏，到 2015 年承担当代多数科学计算，再到 2025 年运行世界最大的认知基础设施。二十年五步。
From a single chip to a hypothetical civilisation-scale fabric. Each rung up multiplies the cognitive scaffolding by an order of magnitude in cost. Each also makes a different class of capability viable that the rung below could not support.从单颗芯片到一个假设的文明级织物。每爬一级，认知脚手架的成本即扩大一个数量级。每爬一级，也使得下一级所无法支撑的一类能力得以成立。
Salaries, funding, prestige, status arbitrage, and what money does not buy. Read together, the five mechanisms describe a system that prices a narrow class of cognitive contributions extremely well — and almost everything else extremely badly.薪资、资助、声望、地位套利，以及钱所买不到者。五种机制合在一起，描绘的是一个对一类窄认知贡献极擅定价、对其他几乎一切极不擅定价的系统。
Cognition keeps leaving the body. Each layer multiplies the previous and never replaces it. Universities still exist; libraries still exist; print still exists; AI is the latest layer, not the only one running.认知不断离开身体。每一层都放大前者，且从不取代之。大学仍在，图书馆仍在，印刷仍在；AI 是最新的一层，并非唯一在运转的一层。
Move the dials to assemble a regime. The verdict at the bottom matches your configuration to a recognisable archetype: concentration regime, mass-cognition regime, talent-without-infrastructure, capital-monopoly, or high-fragility configuration.移动旋钮以组装体制。底部判决会将您的配置匹配至可辨识的原型：集中体制、大众认知体制、有才无基础设施、资本垄断，或高脆弱配置。
Each scenario is partial — none of them is a destination. Read them as orientations: what is already happening, what is plausible, what is speculative. The honest position is to take all four seriously without endorsing any.每一情境都是局部的——无一是终点。请将其视为方向：已发生者、合理推测者、推测性者。诚实立场是认真对待四者，但不承诺任一。
A civilisation that amplifies cognition this fast must build proportionate institutions for steering. The questions below are not new — analogues exist for railways, oil, banking, broadcasting. The answers in those previous chapters were partial, slow, and politically contested. The next decade will write a comparable chapter for cognitive infrastructure.一个以此速度放大认知的文明，必须建立成比例的引导制度。下列问题并非新问——铁路、石油、银行、广播都有类比。先前章节的答案皆局部、缓慢、政治化。下一十年将为认知基础设施书写一个可比的章节。