瓦房店学的历史间隙与瓦房店化平方：中国的科技发展可以暂时取得领先吗?Can China’s Technological Development Ever Take a Temporary Lead?

文章全文

By Timothy Huang

如今，人们通常以“瓦房店学”解释中国科技发展长期落后于世界先进水平的根本原因。一方面，东亚大陆内部长达两千余年的去封建化，使得稳定的技术共同体难以存在，技术知识无法通过职业团体、行业伦理和代际传承实现持续积累。相反，在高度官僚化的体制筛选下，官僚机制经常以短期利益和行政目标为导向，排斥科学技术的自发生长和有效应用。另一方面，中国长期依赖外部技术输入。在后发国家的现实处境中，购买和引进往往比自主研发更具短期效率；技术输入相对于本土原生创新具有无可比拟的成本优势，于是进一步削弱了自主研发和跨代积累的可能性。二者结合，便形成了一种典型现象：外部输入的已经落伍的技术在本地看似“遥遥领先”，但当地自身的科技基础和迭代能力却不断退化。

这套理论当然能够解释中国科技发展为何在根本上长期落后于世界先进水平。然而，它并没有充分解释，或者说并没有细致解释另外三个问题。

第一，为什么中国科技发展不仅从较长历史周期看始终落后于世界先进水平，而且在几乎每一个具体历史节点上，也始终难以达到真正的世界领先水平？即使在世纪之交的全球化黄金时期，中国获得了空前规模的技术引进、产业转移和商业窃取机会，也很难在某个短暂历史节点、某几项特定领域内真正站到世界前沿。

第二，既然依赖外部技术输入并非中国独有，而是几乎所有非欧美核心国家都普遍存在的现象，为什么中国仍然要以相对更低的效率、甚至以数倍于其他依赖外部输入地区的资源消耗，才能实现外部输入技术的平均效果？

第三，基于前两点，为什么在当下中国瞄准AI、电动车、机器人等特定行业，投入数倍乃至数十倍于其他国家的资源，并且已经通过各种合法或非法手段获得接近世界先进水平的技术输入之后，仍然无法在这些领域取得哪怕短暂、局部而真实的领先地位？

我把这三个问题称为“瓦房店学的历史间隙问题”。对这三个问题的解释，则可以称为“再瓦房店化”，或者说“瓦房店化的平方”。

必须承认，瓦房店学提供了一条基本正确的解释路径。可是，当一些人出于各种原因，认为中国在AI、电动车、机器人等宣传极为成功的行业中已经达到了世界先进水平，并据此为中国科技成就欢呼时，瓦房店学的基础版本并不能提供一个实时可用的解释。那些为中国科技发展感到自豪的人，即使愿意在抽象层面接受瓦房店学，也仍然可能认为，当下中国恰好处在瓦房店学历史解释的“间隙”之中：刚刚引入了真正的世界先进技术，同时又利用列宁主义体制集中投入更多资源；即使五年或十年后终究会因为瓦房店化而重新落后，至少不妨碍他们在当下宣称已经胜利。

本文要讨论和驳斥的，正是这种“历史间隙中的赢在当下”叙事。所谓瓦房店化平方，就是要说明：先进技术一旦进入中国，并不会在短期内形成真正的领先状态；它从输入之初就已经发生折损，随后又在组织结构、资源配置和迭代传承中继续退化。因此，中国科技发展并不存在一个可以绕过瓦房店化的短暂窗口。

与许多国家和地区都普遍适用的、单纯因为技术传统无法在本地跨代积累而导致的一般瓦房店化相比，中国之所以会出现瓦房店化的平方，主要有三个原因。

第一，科学技术的产生、发展、传承和进步依赖于封建性的技术共同体；而技术本身的应用和生命力，同样依赖于这一土壤。把一朵鲜花从原本的土壤中摘下来的那一刻，它虽然仍然是花，却已经进入枯萎过程。因此，理论上最先进的科技在输入中国的那一刻，便已经不再是与原生地同等水平的科技。

第二，中国社会结构对科技发展和应用的影响，不能只从封建性长期缺失、技术共同体难以形成的角度理解，还必须从社会资源组织难度的角度理解。中国社会在横向信任、组织稳定性、责任继承和长期合作方面的成本极高，以至于只有匍匐在传统政治文化之下的列宁主义体制，才能在某种程度上实现有效资源配置。然而，这种配置方式本身又会继续破坏技术共同体的生成。

第三，科技在输入过程中的损耗，与中国极高的资源组织难度叠加之后，使得中国科技水平与技术输入水平、资源投入规模之间并不存在简单线性关系。它更像是一种低于科技原生地水平的低位收敛。AI、电动车、机器人等行业即使借助开源技术、技术窃取和不计成本的资源投入，也很难真正达到世界先进水平。

瓦房店化平方的第一个方面，是技术离开原生土壤的那一刻，就已经发生第一次质量折损。

瓦房店学的基础版本通常假设，外部输入的科技本身就不是当时世界最先进的水准，因为将最先进技术转移给后发地区，不符合拥有这些技术的先进地区的自身利益。这个判断当然成立。但它也导致基础版本的瓦房店学难以应对一种极端情形：如果中国通过非常规手段，确实取得了当时世界上最先进的科技理论、成果、设备或工程经验，又该如何解释其仍然不能取得哪怕是暂时的真正领先？这正是必须引入瓦房店化平方的直接原因。

许多中国的民族自豪家们理解科技的方式十分幼稚。它们把科技理解为一套图纸、一条产线、一篇论文、一个模型权重、一个开源代码库、一名被挖来的工程师，或者一批从海外购买的设备。似乎只要这些东西被放进中国的实验室、工厂和产业园，中国就拥有了同等水平的科技。如果这些技术并非别人用剩下的二流技术，而是确实代表了世界顶尖水准，那么基础版本的瓦房店学似乎就无法解释中国为何仍然不能暂时领先。

然而，真正的科技从来不是孤立的成品，而是一个生态。它包括上游材料、基础理论、工艺经验、设备调校、质量控制、工程师共同体、行业标准、客户反馈、法律责任、职业伦理，以及围绕这一切形成的长期稳定预期。它在原生地并非孤立存在，而是像森林里的树一样，与根系、土壤、微生物、水文环境和气候条件相连。树苗被移走以后，能否活下来是一回事；能否长成原来那片森林里的参天大树，是另一回事。鲜花被摘下时依然是花，但从摘下的那一秒起，它已经离开了自己的生命系统。

因此，哪怕中国输入的确实是理论上最先进的科技，不考虑时间差，它进入中国时也已经不是与原产地同一等级的科技，这正是切片与系统的差距。中国获得的永远只是某一个时间切片上的成果，是被拆成零件、说明书、专利文本、源代码和专家履历的碎片。而如果考虑时间因素，原产地正在不断迭代和升级的科技水平，就更加确定地优于在此之前某个时间节点被中国引进甚至是盗窃而得的科技水平了。技术刚被输入的时候，就已经从活体技术变成了标本技术。活体技术可以自我修复、自我繁殖、自我进化；标本技术只能展示、模仿和解剖。

因此，外部输入的科技本身不可能真正保持当时世界最先进水准，不仅因为现实层面上这不符合先进科技拥有者的自身利益，也因为从逻辑上讲，外部输入的科技一旦离开原生生态，就已经不再具有完整意义上的先进性。

有人或许会问，难道某些先进科技不能恰好在中国遇到更适合的应用生态，从而维持领先水平，甚至更上一层楼吗？这种疑问实际上混淆了“应用生态”和“技术原生生态”。某些先进科技进入中国以后，当然可能遇到更大的市场、更低的监管阻力、更便宜的工程师、更完整的制造供应链、更激进的资本，以及更愿意配合产业政策的地方政府。这些条件适合扩散、降本、包装、量产和商业化，却并不适合源头创新、技术传统、核心工具链和下一代范式定义。

换言之，中国确实可能成为某些技术的涡轮增压器，但涡轮增压器和发动机并不是同一回事。增压器可以提高短期输出，可外部输入的发动机本身才是内核；而不当改装的增压器，往往还会加剧发动机自身的水土不服。瓦房店化平方并不否认中国在应用层可能短期放大先进技术的效果；它要指出的是，先进技术进入中国以后，会在源头层、组织层、传承层和迭代层发生折损。中国所谓“更合适的生态”，多半适合技术的消费和扩张，却不适合技术的自治和传承。前者可以制造产业爆发，后者才能制造技术传统。

我们都非常熟悉典型的中国式技术引进。它们往往在宣传阶段声势巨大，落地阶段开始降级，量产阶段开始变形，维护阶段开始敷衍，下一代迭代阶段又重新归零。这也解释了为什么中国特别擅长制造“看起来像”的东西。看起来像高铁，看起来像芯片产业，看起来像AI大模型，看起来像新能源车，看起来像人形机器人，看起来像现代大学，看起来像世界级实验室。它们确实可以很像，甚至在某些技术指标上实现超越，因为外观、流程、指标和商业包装都可以被复制。但“像”这个字本身已经暴露了问题。真正的世界领先水平并不需要证明自己像谁，因为它本身就是别人模仿的对象。中国科技宣传中最常见的句式，恰恰是“国产某某媲美某某”“中国某某追平某某”“某某领域打破西方垄断”，以及更常见的“遥遥领先”。这种宣传语言本身就说明，其参照系仍然在外部。

一个现实例子是黄仁勋近来不断游说美国政府要求对中国解禁先进制程芯片出口。他的理由是，如果美国不卖，中国就会自己做；与其如此，不如让中国继续依赖美国技术，同时让美国企业获得收益。这个判断中的前提并不稳固。如果美国不向中国出口先进制程芯片，中国AI发展与美国的差距显然会继续扩大，而不是迅速形成一套可以与美国竞争的独立体系。

但反过来说，批评黄仁勋的人如果认为，只要美国向中国出售先进制程芯片，中国AI就会立刻迎头赶上，这同样夸大了技术输入的作用。中国AI的确在论文、专利、应用场景和模型表现上取得了明显进步。但中国AI的强项主要体现在追赶速度、应用规模和工程压缩能力上，更接近于在既有范式下迅速做题、跑分和堆数据。最前沿范式的定义权、算力生态、底层芯片和核心工具链，仍然高度受制于外部技术中心。英伟达芯片的技术指标放在中国和美国或许相同，甚至由于中国在某些地区拥有相对稳定和低廉的电力供应，实际运营成本可能略有优势。然而，科技发展并不是单纯比较跑分。即使把中文语料喂给最新一代架构训练，实际效果也未必优于上一代架构下的成熟英文生态，而这正是中国三千年来不断把文字当作刘仲敬所说的“用金圆券换金子的把戏”的恶果。中国式叙事只需要一个速度、规模或成本指标，就足以宣布胜利；但真正的技术领先并不由单一指标决定。更何况，即使出口限制放宽，中国也未必能获得真正最前沿的一代产品。两代之前的硬件即使通过各种灰色渠道进入中国，中国自己的AI生态最终发展成什么样，仍然要看底层工具链、开发者网络、原创范式和组织结构，而不只是硬件本身。

所以，那些被中国赢学家反复援引的“先进科技输入中国以后反而更先进”的案例，恰恰证明了瓦房店化平方：低成本、强动员、高扩张的洼地作用，并没有真正提升技术原生能力，反而可能加速先进科技的扭曲、透支和退化。

瓦房店化平方的第二个方面，是中国社会组织资源的成本高到反常。任何先进技术进入中国之后，无论是技术本身的应用，还是进一步国产化，都要先被组织结构消耗一遍。

一般后发国家的问题，是缺乏原生技术传统，所以必须引进、学习、消化和吸收。按照瓦房店学的一般原理，本土科研工作者再怎么努力，也会被外部引进的先进科技以低成本和高质量压倒，于是本土很难产生有效的科技发展和积累；反过来，如果坚持扶植本土科研，就必须投入远高于直接引进的资源。这一基础版本的瓦房店学，在比较本土科技和外来输入科技的资源消耗时，主要考虑的是先进地区与落后地区之间不同的科技基础，以及技术共同体相关的社会组织结构差异。这是所有非科技核心区共同面对的劣势，许多中国民族自豪叙事也未必完全否认这一点。

然而，中国的问题更深一层：它连“有效学习”本身都需要付出极高成本。技术吸收并不是一个人看懂说明书，也不是一家企业买到设备，而是大量主体之间长期配合的结果。高校要能够培养可信的人才，企业要愿意做长期研发，供应商要稳定交付，客户要提出高质量需求，法院要保护合同，资本市场要奖励长期主义，行业协会要形成标准，工程师要在职业共同体中积累声誉。所有这些，都要求社会中存在较高水平的横向信任、组织稳定性和责任继承机制，而不仅仅是狭义意义上的技术共同体。中国社会恰恰在这些方面成本极高。人与人之间、企业与企业之间、地方与地方之间，普遍缺乏稳定、可预期和可继承的合作关系。血缘、权力、行政层级和短期利益，经常成为最可靠甚至仅有的组织纽带。任何需要长期积累的技术事业，一进入这样的环境，就会首先被切割成项目、指标、政绩、融资故事和宣传材料。

因此，中国式资源投入，看起来是倾全国之力，实际却经历层层损耗。中央提出突破“卡脖子”，部委开始立项，地方开始建园区，国企开始投资，民企开始讲故事，学校开始设学院，媒体开始造神，资本市场开始炒概念。从资源投入到真实技术能力之间，要经过一个漫长的漏斗。每一层都可能截留，每一层都可能改写，每一层都需要表演，每一层都要把复杂的技术问题翻译成上级能够理解的政绩语言。最后真正沉淀到工程能力中的部分，可能只是原始投入的一小部分。

这正是列宁主义在中国具有特殊诱惑力的原因。在组织度极低的社会中，自发秩序难以维持，于是不得不依靠列宁主义国家机器实现最大规模的资源投入，希望在层层损耗后仍然留下尽可能多的实际成果。也只有列宁主义国家机器可以违背市场规律和自发秩序，把一盘散沙强行压成砖，把彼此不信任的主体依靠强制力塞进同一个工程之中，并在名义上将其合成一股力量。然而，这种行为本身又会破坏技术共同体。技术共同体依赖声誉、专业判断和长期责任；列宁主义动员依赖服从、考核、排名和政治安全。前者培养人，后者消耗人；前者让知识在共同体中生长，后者让知识在项目中燃烧。前者形成传统，后者制造运动。

况且，中国的列宁主义体制并不是原汁原味的苏联进口货，而是与中国传统文化结合后的产物。这种结合既削弱了计划经济在技术管理上的专业性，也削弱了特务体系管控社会的高效性，更削弱了列宁主义通过军事征服和国际主义动员从外部汲取资源的能力。中国体制的问题，恰恰在于它经常同时继承了传统文化和列宁主义二者中最不利于技术共同体的一面。

所以，中国科技投入的问题不仅在于基础版本瓦房店学所指出的那样：本土研发在效率上不如外部引进，因而常常半途而废。更重要的是，它还受制于中国社会结构本身，进一步降低了后发国家原本就很低的自主科研资源使用效率。这种社会结构的危机，也远不止去封建化和官僚体系导致技术共同体消亡那么简单，而是整个社会在传统政治文化与列宁主义组织逻辑的叠加下持续内耗。中国赢学家和民族自豪叙事往往只关注投入到这个体系之前的资源总量，却忽略了这些资源在层层过滤网中被消耗、扭曲和重写的过程。也正因为如此，他们才会得出“中国可以靠大力出奇迹的体制优势逆转瓦房店化”的错误结论。

瓦房店化平方的第三个方面，是技术输入损耗和资源组织损耗叠加以后，中国科技水平与技术输入、资源投入之间形成一种低位收敛关系。

很多人以为，只要中国获得足够先进的技术，再投入足够多的钱，总有一天就能追上世界第一。这个想法的错误在于，它默认技术输入和资源投入可以线性转化为自有技术水平。似乎世界先进水平是一百分，中国输入八十分技术，再投入两倍资源，就可以得到一百六十分。

瓦房店化平方说明，这一关系并不是线性的。中国的实际科技水平更接近于这样一个公式：实际技术水平，等于输入技术水平，乘以土壤保真率，再乘以组织转化率，最后再扣除迭代滞后。

如果第一部分所说的土壤保真率小于一，第二部分所说的组织转化率也小于一，那么输入九十分技术，最后可能只剩六十分；投入十倍资源，经过组织漏斗以后，可能只剩两三分有效投入；用五年追赶，原产地已经又向前走了五年。于是，中国科技发展经常出现一种悖论：越是重点突破的行业，越容易产生巨大的宣传泡沫；越是投入巨大的领域，越容易形成低效重复建设；越是短期内看起来热闹，越容易在下一轮技术范式变化中重新暴露根基薄弱。

这就是为什么AI、电动车和机器人这几个行业特别适合用来解释瓦房店化平方。它们都有共同特征：外部技术输入极其丰富；开源、供应链和全球化分工降低了追赶门槛；中国政府和资本愿意不计成本地投入；应用场景庞大，商业化速度很快；宣传效果极好。按照民族自豪叙事，这些行业最应该证明中国已经可以绕开瓦房店，至少可以赢在当下。可恰恰是这些行业，最能说明“赢在当下”的脆弱性。AI可以拿开源模型，可以购买算力，可以吸收人才，可以堆数据，可以利用庞大应用场景快速打磨产品。但一旦问题进入最前沿模型架构、先进芯片、训练集质量、基础软件生态、全球开发者网络和原创范式，中国的追赶就会重新面对外部技术中心。电动车可以依靠规模、供应链、价格和政策打出世界级产业，却仍然要面对利润率、海外监管、品牌信任、智能驾驶安全、核心零部件可靠性和地缘政治阻断。机器人可以依靠巨大制造业市场堆出全球最大装机量，却仍然要面对高端零部件、复杂场景泛化、软件智能、可靠性和真正人机协作能力的考验。中国可以在某一个指标上非常亮眼，可以在某一个市场形成压倒性规模，可以在某一个产品阶段把价格打到令全世界恐惧。但这并不自动转化为世界先进水平。市场占有率既可能意味着真正的技术吸引力，也可能意味着廉价、补贴、低利润和外部环境尚未设防。单项指标、短期规模和价格优势，并不能直接证明技术体系本身已经具备原生领先能力。所谓领先，往往是用局部指标冒充体系领先，用商业规模冒充技术原生性，用补贴产能冒充创新能力，用开源追赶冒充范式定义，用短期爆款冒充长期传统。

综上所述，瓦房店化平方意味着：先进技术进入洼地的第一天，就已经因为脱离土壤而降级；在应用过程中，又因为组织成本过高而再次降级；在追赶过程中，又因为原产地持续迭代而永远追不上。它不是五年之后才落后，也不是十年之后才落后；而是在宣布“遥遥领先”的那一刻，领先叙事本身就已经落后于真实技术生态。

瓦房店化不存在真正的历史间隙。在瓦房店化平方理论之下，中国科技在任何时间点、任何规模的资源投入下，都很难在真正意义上达到世界领先水平。中国科技试图绕过瓦房店化的常见方式，就是不断把局部成果包装成整体突破。它不能稳定定义技术路线，不能形成独立而持久的职业共同体，不能跨代积累默会知识，不能在没有外部输入的情况下继续迭代。于是，它只能一次又一次回到瓦房店化平方的收敛曲线之中。民族自豪家们所谓“在瓦房店化的历史间隙里赢在当下”，不过是在技术输入尚未完全枯竭、财政补贴尚未完全耗尽、外部市场尚未完全设防、原产地下一代技术尚未完全展开之前，站在一座临时搭建的舞台上高喊胜利。等灯光熄灭，布景拆掉，后台显露出来，人们就会发现，舞台下面依然是瓦房店。

Wafangdianisation Squared and the Historical Gap in Liu Zhongjing’s Wafangdian Studies: Can China’s Technological Development Ever Take a Temporary Lead?

By Timothy Huang

Today, people commonly use Liu Zhongjing’s ‘Wafangdian Studies’ to explain the fundamental reason why China’s technological development has long lagged behind the world’s advanced level. On the one hand, more than two thousand years of de-feudalisation within the East Asian mainland made it difficult for stable technical communities to exist, and technical knowledge could not be accumulated continuously through professional associations, industry ethics and intergenerational transmission. On the contrary, under the screening of a highly bureaucratised system, bureaucratic mechanisms have often been guided by short-term interests and administrative objectives, thereby excluding the spontaneous growth and effective application of science and technology. On the other hand, China has long relied on external technological input. In the real circumstances of late-developing countries, purchasing and importing technology often has greater short-term efficiency than independent research and development; technological input possesses an incomparable cost advantage over indigenous innovation, and therefore further weakens the possibility of autonomous research and intergenerational accumulation. The combination of the two has produced a typical phenomenon: already outdated technology imported from outside appears locally to be ‘far ahead’, while the local technological foundation and capacity for iteration continue to deteriorate.

This theory can of course explain why China’s technological development has, at root, long lagged behind the world’s advanced level. Yet it does not fully explain, or at least does not explain in sufficient detail, three further questions.

First, why has China’s technological development not only remained behind the world’s advanced level over long historical cycles, but also struggled, at almost every specific historical moment, to reach a genuinely world-leading level? Even during the golden age of globalisation around the turn of the century, when China obtained opportunities for technological importation, industrial transfer and commercial theft on an unprecedented scale, it still found it difficult to stand at the global frontier, even in a few specific fields and even for a brief historical moment.

Secondly, since dependence on external technological input is not unique to China, but is a phenomenon commonly found in almost all countries outside the Euro-American core, why does China still have to use relatively lower efficiency, and even resources several times greater than those consumed by other regions dependent on external input, merely to achieve the average effect of imported technology?

Thirdly, on the basis of the previous two points, why is it that China today, after targeting particular industries such as AI, electric vehicles and robotics, investing several times or even several dozen times more resources than other countries, and already obtaining technological inputs approaching the world’s advanced level through various lawful or unlawful means, still cannot achieve even a temporary or partial genuine lead in these fields?

I call these three questions the ‘historical gap problem’ of Wafangdian Studies. The explanation of these three questions may be called ‘re-Wafangdianisation’, or ‘Wafangdianisation squared’.

It must be admitted that Wafangdian Studies provides a basically correct explanatory path. Yet when, for various reasons, some people believe that China has already reached the world’s advanced level in heavily promoted industries such as AI, electric vehicles and robotics, and cheer China’s technological achievements on that basis, the basic version of Wafangdian Studies does not provide an immediately usable explanation. Those who take pride in China’s technological development, even if they are willing to accept Wafangdian Studies at an abstract level, may still believe that China today happens to be situated in a ‘gap’ within the historical explanation offered by Wafangdian Studies: it has just imported genuinely world-advanced technology, while simultaneously using a Leninist system to concentrate greater resources. Even if China is ultimately doomed to fall behind again because of Wafangdianisation in five or ten years, this at least does not prevent them from declaring victory in the present.

What this article seeks to discuss and refute is precisely this narrative of ‘winning in the present within the historical gap’. Wafangdianisation squared is intended to show that once advanced technology enters China, it does not form a genuinely leading position even in the short term. From the very beginning of its importation, it has already suffered degradation; thereafter it continues to deteriorate in organisational structure, resource allocation and iterative inheritance. China’s technological development therefore contains no brief window in which Wafangdianisation can be bypassed.

Compared with the general form of Wafangdianisation, which applies widely to many countries and regions and results simply from the inability of a technological tradition to accumulate locally across generations, the reason China experiences Wafangdianisation squared lies mainly in three factors.

First, the emergence, development, transmission and progress of science and technology depend on feudal technical communities; the application and vitality of technology itself likewise depend on this soil. The moment a flower is picked from its original soil, it is still a flower, but it has already entered the process of withering. Therefore, at the moment when theoretically most advanced technology is imported into China, it is already no longer technology of the same level as that of its place of origin.

Secondly, the impact of China’s social structure on technological development and application cannot be understood only from the perspective of the long-term absence of feudalism and the difficulty of forming technical communities. It must also be understood from the perspective of the difficulty of organising social resources. The cost of horizontal trust, organisational stability, inherited responsibility and long-term cooperation in Chinese society is extremely high, to the point that only a Leninist system crouched beneath traditional political culture can, to some extent, achieve effective resource allocation. Yet this mode of allocation itself continues to undermine the formation of technical communities.

Thirdly, once the loss incurred in the process of technological importation is combined with China’s extremely high difficulty in organising resources, there is no simple linear relationship between China’s technological level, the level of technological input and the scale of resource investment. It more closely resembles a low-level convergence below the level of the technology’s place of origin. Even with open-source technology, technological theft and resource investment regardless of cost, industries such as AI, electric vehicles and robotics still find it difficult to reach a genuinely world-leading level.

The first aspect of Wafangdianisation squared is that, at the very moment technology leaves its native soil, it has already undergone its first degradation in quality.

The basic version of Wafangdian Studies usually assumes that externally imported technology is not itself the most advanced level in the world at that time, because transferring the most advanced technology to late-developing regions does not serve the interests of the advanced regions that possess it. This judgement is of course correct. Yet it also makes the basic version of Wafangdian Studies struggle to respond to an extreme scenario: if China, through irregular means, does indeed acquire the most advanced scientific theories, achievements, equipment or engineering experience in the world at that time, how should one explain the fact that it still cannot obtain even a temporary genuine lead? This is the direct reason why Wafangdianisation squared must be introduced.

Many Chinese national-pride enthusiasts understand technology in an extremely childish way. They treat technology as a set of drawings, a production line, a paper, a set of model weights, an open-source code repository, a poached engineer, or a batch of equipment purchased from overseas. It is as if, once these things are placed in Chinese laboratories, factories and industrial parks, China possesses technology of the same level. If these technologies are not second-rate leftovers from others, but do indeed represent the world’s top level, then the basic version of Wafangdian Studies would appear unable to explain why China still cannot take even a temporary lead.

Yet real technology is never an isolated finished product, but an ecosystem. It includes upstream materials, basic theory, process experience, equipment tuning, quality control, engineering communities, industry standards, customer feedback, legal responsibility, professional ethics, and the long-term stable expectations formed around all of these. In its place of origin, it does not exist in isolation, but resembles a tree in a forest, connected to its roots, soil, microorganisms, hydrological environment and climatic conditions. Whether a sapling can survive after being moved is one question; whether it can grow into a towering tree like those in the original forest is another. A flower is still a flower when it is picked, but from the second it is picked, it has already left its own life system.

Therefore, even if the technology China imports is, in theory, the most advanced, and even setting aside any time lag, it is already not technology of the same grade as that in its place of origin when it enters China. This is precisely the gap between a slice and a system. What China obtains is always the result of a particular time slice: fragments broken down into components, manuals, patent texts, source code and expert CVs. If the time factor is taken into account, the technological level of the place of origin, which is continuously iterating and upgrading, is even more certainly superior to the technological level imported or even stolen by China at some earlier point in time. The moment technology is imported, it has already changed from living technology into specimen technology. Living technology can repair itself, reproduce itself and evolve itself; specimen technology can only be displayed, imitated and dissected.

For this reason, externally imported technology cannot truly maintain the most advanced level in the world at the time. This is not only because, at the practical level, doing so does not serve the interests of the holders of advanced technology, but also because, as a matter of logic, once externally imported technology leaves its native ecosystem, it no longer possesses advancedness in the complete sense.

Someone may ask whether certain advanced technologies might just happen to encounter a more suitable application ecosystem in China, thereby maintaining a leading level or even advancing further. This question in fact confuses two different levels: the ‘application ecosystem’ and the ‘native technological ecosystem’. After entering China, certain advanced technologies may of course encounter a larger market, lower regulatory resistance, cheaper engineers, a more complete manufacturing supply chain, more aggressive capital and local governments more willing to cooperate with industrial policy. These conditions are suitable for diffusion, cost reduction, packaging, mass production and commercialisation, but they are not suitable for source innovation, technological tradition, core toolchains or the definition of the next paradigm.

In other words, China may indeed become a turbocharger for certain technologies, but a turbocharger and an engine are not the same thing. A turbocharger can increase short-term output, but the externally imported engine itself remains the core; moreover, an improperly modified turbocharger will often exacerbate the engine’s own maladaptation to its environment. Wafangdianisation squared does not deny that China may temporarily amplify the effect of advanced technologies at the application level. What it seeks to point out is that, after entering China, advanced technologies suffer losses at the level of source, organisation, transmission and iteration. What China calls a ‘more suitable ecosystem’ is generally suited to the consumption and expansion of technology, but not to the autonomy and inheritance of technology. The former can generate industrial explosions; the latter alone can generate technological traditions.

We are all familiar with the typical Chinese-style introduction of technology. It often creates enormous momentum during the propaganda stage, begins to degrade during implementation, begins to deform during mass production, becomes perfunctory during maintenance, and returns to zero again at the stage of next-generation iteration. This also explains why China is particularly adept at producing things that ‘look like’ something else. It looks like high-speed rail, looks like a chip industry, looks like an AI, looks like new-energy vehicles, looks like humanoid robots, looks like modern universities, and looks like world-class laboratories. They can indeed look very similar, and may even surpass others on certain technical indicators, because appearance, process, indicators and commercial packaging can all be copied. Yet the word ‘like’ itself already exposes the problem. A truly world-leading level does not need to prove that it resembles anyone else, because it is itself the object others imitate. The most common sentence patterns in Chinese technological propaganda are precisely ‘domestic X rivals Y’, ‘China’s X has caught up with Y’, ‘a certain field has broken the Western monopoly’, and, more commonly, ‘far ahead’. This propaganda language itself shows that its frame of reference remains external.

A practical example is Jensen Huang’s recent lobbying of the US Government to lift restrictions on the export of advanced-process chips to China. His reasoning is that, if America does not sell them, China will make them itself; if so, it would be better to keep China dependent on American technology while allowing American companies to profit. The premise of this judgement is not solid. If the United States does not export advanced-process chips to China, the gap between China’s AI development and that of the United States will clearly continue to widen, rather than China rapidly forming an independent system capable of competing with America.

Conversely, however, if Jensen Huang’s critics believe that selling advanced-process chips to China would immediately allow Chinese AI to catch up, that too exaggerates the role of technological input. Chinese AI has indeed made obvious progress in papers, patents, application scenarios and model performance. Yet China’s strengths in AI lie chiefly in catch-up speed, application scale and engineering compression capability. It is closer to rapidly solving problems, running benchmarks and piling up data within an existing paradigm. The power to define the most advanced paradigms, the computing ecosystem, the underlying chips and the core toolchains remain highly dependent on external technological centres. Nvidia chips may have the same technical specifications in China and in the United States, and because some regions in China have relatively stable and cheap electricity supplies, actual operating costs may even be slightly advantageous. Yet technological development is not simply a comparison of benchmark scores. For example, even if Chinese-language data are fed into the latest-generation architecture for training, the actual effect may not surpass a mature English-language ecosystem built on a previous-generation architecture. This is precisely the consequence of China’s three-thousand-year habit of treating writing as what Liu Zhongjing called ‘the trick of exchanging gold-yuan notes for gold’. A Chinese-style narrative needs only a single metric of speed, scale or cost to declare victory; genuine technological leadership is not determined by a single metric. What is more, even if export restrictions are relaxed, China may still not obtain the truly most advanced generation of products. Even if hardware from two generations ago enters China through various grey channels, the eventual state of China’s own AI ecosystem will still depend on underlying toolchains, developer networks, original paradigms and organisational structures, not merely on the hardware itself.

Therefore, those cases repeatedly cited by Chinese victory theorists in which ‘advanced technologies become even more advanced after entering China’ in fact prove Wafangdianisation squared: the low-cost, high-mobilisation and high-expansion effects of the lowland do not truly enhance indigenous technological capacity, but may instead accelerate the distortion, overdrawn use and degradation of advanced technology.

The second aspect of Wafangdianisation squared is that the cost of organising resources in Chinese society is abnormally high. After any advanced technology enters China, whether in the application of the technology itself or in its further localisation, it must first be consumed by the organisational structure.

The problem faced by ordinary late-developing countries is that they lack an indigenous technological tradition, and therefore must import, learn, digest and absorb. According to the general principle of Wafangdian Studies, no matter how hard local researchers work, they will be overwhelmed by externally imported advanced technology with lower cost and higher quality. Local society therefore finds it difficult to generate effective technological development and accumulation; conversely, if it insists on supporting local research, it must invest far more resources than direct importation would require. When comparing the resource consumption of local technology and imported external technology, this basic version of Wafangdian Studies mainly considers differences in technological foundations between advanced and backward regions, as well as differences in social organisational structures related to technical communities. This is a common disadvantage faced by all non-core technological regions, and many Chinese national-pride enthusiasts may not entirely deny it.

Yet China’s problem goes deeper: even ‘effective learning’ itself requires an extremely high cost. Technological absorption is not a matter of one person understanding a manual, nor of one company purchasing equipment. It is the result of long-term coordination among many actors. Universities must be able to train trustworthy talent, enterprises must be willing to undertake long-term research and development, suppliers must deliver steadily, customers must put forward high-quality demands, courts must protect contracts, capital markets must reward long-termism, industry associations must form standards, and engineers must accumulate reputations within professional communities. All of these require a relatively high level of horizontal trust, organisational stability and mechanisms for inheriting responsibility within society, not merely a technical community in the narrow sense. Chinese society has extremely high costs precisely in these respects. Between individuals, between enterprises and between localities, stable, predictable and inheritable cooperative relationships are generally lacking. Blood ties, power, administrative hierarchy and short-term interests often become the most reliable, or even the only, organisational bonds. Any technological undertaking requiring long-term accumulation, once it enters such an environment, is first cut up into projects, indicators, political achievements, financing stories and propaganda materials.

Thus Chinese-style resource investment appears to mobilise the strength of the whole country, but in reality it passes through layer upon layer of loss. The centre proposes a breakthrough in ‘chokepoint’ technologies; ministries begin setting up projects; localities begin building industrial parks; state-owned enterprises begin investing; private firms begin telling stories; universities begin establishing new schools; the media begin creating heroes; the capital market begins speculating on concepts. Between resource input and real technological capability lies a long funnel. Every layer may intercept resources, every layer may rewrite objectives, every layer needs performance, and every layer must translate complex technical problems into the language of political achievement that superiors can understand. The part that finally settles into engineering capacity may be only a small fraction of the original input.

This is precisely why Leninism has a special appeal in China. In a society with an extremely low degree of organisation, spontaneous order is difficult to sustain. It therefore becomes necessary to rely on the Leninist state machine to achieve the largest possible scale of resource input, in the hope that, after layer upon layer of loss, as much real output as possible will remain. Only the Leninist state machine can violate market laws and spontaneous order, forcibly press loose sand into bricks, force mutually distrustful actors into the same project through coercive power, and nominally combine them into a single force. Yet this behaviour itself also destroys technical communities. Technical communities rely on reputation, professional judgement and long-term responsibility; Leninist mobilisation relies on obedience, assessment, ranking and political security. The former cultivates people; the latter consumes people. The former allows knowledge to grow within communities; the latter burns knowledge inside projects. The former forms traditions; the latter manufactures campaigns.

Moreover, China’s Leninist system is not an unadulterated Soviet import, but a product of its combination with traditional Chinese culture. This combination has weakened the professionalism of the planned economy in technological management, weakened the efficiency of the secret-police system in controlling society, and further weakened the Leninist ability to draw resources from outside through military conquest and internationalist mobilisation. The problem with the Chinese system is precisely that it often inherits simultaneously those aspects of traditional culture and Leninism most unfavourable to technical communities.

Therefore, the problem of Chinese technological investment does not lie only in what the basic version of Wafangdian Studies identifies: local research and development is less efficient than external importation, and therefore often ends halfway. More importantly, it is also constrained by China’s social structure itself, which further reduces the already low efficiency with which late-developing countries use resources for autonomous research. The crisis of this social structure also goes far beyond the disappearance of technical communities caused by de-feudalisation and bureaucratic systems. It is a continuing internal attrition of the entire society under the combination of traditional political culture and Leninist organisational logic. Chinese victory theorists and national-pride enthusiasts often focus only on the total quantity of resources before they enter this system, while ignoring the process by which those resources are consumed, distorted and rewritten in layer upon layer of filters. Precisely for this reason, they arrive at the mistaken conclusion that China can reverse Wafangdianisation by relying on the institutional advantage of producing miracles through sheer force.

The third aspect of Wafangdianisation squared is that, after the losses from technological input and the losses from resource organisation are combined, a low-level convergence relationship forms between China’s technological level, technological input and resource investment.

Many people assume that, so long as China obtains sufficiently advanced technology and invests enough money, it will one day catch up with the world number one. The error in this idea is that it assumes technological input and resource investment can be linearly transformed into indigenous technological capacity. It is as if the world’s advanced level is 100 points, China imports technology worth 80 points, then invests twice the resources and obtains 160 points.

Wafangdianisation squared shows that this relationship is not linear. China’s actual technological level is closer to the following formula: actual technological level equals input technological level, multiplied by the soil fidelity rate, multiplied by the organisational conversion rate, with iterative lag then deducted.

If the soil fidelity rate discussed in the first part is less than one, and the organisational conversion rate discussed in the second part is also less than one, then importing 90-point technology may leave only 60 points in the end; investing ten times the resources, after passing through the organisational funnel, may leave only twenty per cent of effective input; and during the five years spent catching up, the place of origin has already moved another five years ahead. Thus China’s technological development often produces a paradox: the more an industry is designated for priority breakthrough, the more likely it is to generate a huge propaganda bubble; the greater the investment, the more likely it is to produce inefficient and repetitive construction; the livelier things look in the short term, the more easily the weakness of the foundations is exposed again in the next change of technological paradigm.

This is why AI, EVs and robotics are especially suitable for explaining Wafangdianisation squared. They share common features: external technological input is extremely abundant; open source, supply chains and globalised division of labour have lowered the threshold for catching up; the Chinese Government and capital are willing to invest regardless of cost; application scenarios are vast, commercialisation is rapid, and the propaganda effect is excellent. According to the national-pride enthusiasts, these industries should best prove that China has already bypassed Wafangdianisation, or can at least win in the present. Yet precisely these industries best demonstrate the fragility of ‘winning in the present’. AI can use open-source models, purchase computing power, absorb talent, pile up data and rapidly polish products through vast application scenarios. But once the issue enters the realm of frontier model architectures, advanced chips, training-set quality, basic software ecosystems, global developer networks and original paradigms, China’s catching-up once again faces external technological centres. Electric vehicles can rely on scale, supply chains, price and policy to build a world-class industry, yet still have to confront profit margins, overseas regulation, brand trust, intelligent-driving safety, reliability of core components and geopolitical obstruction. Robotics can rely on the enormous manufacturing market to build the world’s largest installed base, yet still faces tests in high-end components, generalisation across complex scenarios, software intelligence, reliability and genuine human-machine collaboration. China can perform brilliantly on one indicator, form an overwhelming scale in one market, and push prices in one product phase down to a level that frightens the world. But this does not automatically translate into a world-leading level. Market share may indicate genuine technological attraction, or it may indicate cheapness, subsidies, low profit margins and an external environment that has not yet built defences. Single indicators, short-term scale and price advantages cannot directly prove that the technological system itself already possesses indigenous leading capacity. So-called leadership often consists of using localised indicators to impersonate systemic leadership, commercial scale to impersonate technological originality, subsidised capacity to impersonate innovative capacity, open-source catching-up to impersonate paradigm definition, and short-term hit products to impersonate long-term tradition.

In summary, Wafangdianisation squared means that, on the first day advanced technology enters the lowland, it has already been downgraded by leaving its soil; during application, it is downgraded again by excessively high organisational costs; and during the process of catching up, it can never catch up because the place of origin continues to iterate. It is not five years later that it falls behind, nor ten years later. Rather, at the very moment when it announces that it is ‘far ahead’, the narrative of leadership itself has already fallen behind the real technological ecosystem.

There is no genuine historical gap in Wafangdianisation. Under the theory of Wafangdianisation squared, Chinese technology, at any point in time and under any scale of resource input, can hardly reach a genuinely world-leading level. The common way in which Chinese technology attempts to bypass Wafangdianisation is constantly to package partial achievements as overall breakthroughs. It cannot stably define technological routes, cannot form independent and durable professional communities, cannot accumulate tacit knowledge across generations, and cannot continue to iterate without external input. It can only return, again and again, to the convergence curve of Wafangdianisation squared. What national-pride enthusiasts call ‘winning in the present within the historical gap of Wafangdianisation’ is merely shouting victory from a temporarily constructed stage before technological input has been fully exhausted, before fiscal subsidies have been fully depleted, before external markets have fully erected defences, and before the next generation of technology from the place of origin has fully unfolded. When the lights go out, the scenery is dismantled and the backstage is revealed, people will discover that beneath the stage, Wafangdian is still there.