把抄袭当成道德污点，就只能靠事后追责；把它当成一种可分析的发生过程，才能看清它在哪一道工序上必然发生——而那道工序，恰恰是查重软件永远看不见的地方。
Treat plagiarism as a moral stain and you can only punish it afterwards. Treat it as an analysable process and you can see the exact step at which it becomes inevitable — the step no similarity checker can see.
洗稿的本体论：论被剽窃的从来不是字句，而是生成理解的那条路径The Ontology of Text-Spinning: What Is Stolen Was Never the Words but the Path That Generated the Understanding
洗稿系统地扔掉字句、保留骨架——而字句正是查重唯一测量的东西，骨架正是抄袭唯一该被指控的东西。于是它证明：被剽窃的从来不是字句，是生成理解的那条论证路径；查重对洗稿的失效不是灵敏度不足，而是构造性的——它测表层，剽窃却在字句所不在的层面。改掉每一个字，仍可原样搬走整条路。
Text-spinning systematically discards the words and keeps the skeleton — yet words are all detection measures, and the skeleton is all plagiarism should indict. Its failure is structural, not technical: the theft happens where the words are not.
读全文 →Read → 王德生 著 · 约 1.1 万字 · 之九by Wang Desheng · ~11k words无源之袭：AI 代写如何暴露“抄袭”从来不是复制问题Sourceless Plagiarism: How AI Ghostwriting Reveals Plagiarism Was Never About Copying
AI 写出的文字没有一个可被侵占的在先作者——按“复制”的旧定义，它既查不出、也算不上抄袭。这恰恰暴露了：抄袭的要害从来不是复制了谁，而是署名与一次真实理解之间的连接被斩断。查重之所以在它面前失效，不是因为不够灵敏，而是它被造来追捕一个被盗的来源，而这里没有来源。
AI-written text has no prior author to rob — by the copying definition it is neither detectable nor plagiarism. That exposes the real object: not what was copied, but a signature uncoupled from any genuine act of understanding.
读全文 →Read → 王德生 著 · 约 2.4 万字by Wang Desheng · ~24k words看到那个焊死的连接：论“产品—主体”连接不被检验的制度生成与再生产Seeing the Welded Joint: How the Unchecked Product–Subject Link Is Institutionally Produced
抄袭不是被道德放行的，是被一道没人负责去问的程序缝隙放行的。匿名评审只在一个方向上真正匿名——审稿人不知道作者；更致命的另一个方向是：审稿人无法知道作者是否真的做过。
Plagiarism is not waved through by morals but by a procedural gap nobody owns. Blind review is blind in only one direction — and no stage is ever assigned the question of whether the signer did the work.
读全文 →Read → 王德生 著 · 约 2.5 万字by Wang Desheng · ~25k words诚实不可计算：论抄袭治理如何拆毁诚实赖以生存的制度条件Honesty Is Not Computable: How Plagiarism Governance Dismantles the Conditions Honesty Needs
查重系统抓到的“抄袭减少”，其实是规避能力的集体增长——可数据看上去一切在变好。市场靠制度的刚性阈值维持交易，制度靠市场产出的漂亮数据维持“治理有效”的合法性：这不是可以修补的漏洞，是双向锁定的共生。诚实的学生反而更可疑，因为他老实引了原始文献。
What the checker records as less plagiarism is a collective rise in evasion skill — yet the data looks like improvement. The market needs the threshold; the institution needs the market's flattering numbers. Not a leak to patch: a symbiosis.
读全文 →Read → 王德生 著 · 约 2.5 万字by Wang Desheng · ~25k words标记空转——论原创性如何被它自身的认证系统所架空The Idling of the Marker: How Originality Is Hollowed Out by Its Own Certification System
原创伦理、量化考核、查重技术，三样东西各有各的历史轨道，却被强行耦合进同一个评价流程。装置在跑，跑得越来越精密，但它咬不住它所标定的对象。洗稿写手与查重工程师看似对立，其实共用同一套语法——字符串的匹配与绕过；谁也没跳出去问一句：论证结构整个搬来、只把句子重写一遍，这算完成了什么原创认知？
Originality ethics, metrics and similarity checking arrived on separate historical tracks and were forced into one pipeline. The device runs, ever more precisely, and never bites its object.
读全文 →Read → 王德生 著 · 约 3.2 万字by Wang Desheng · ~32k words装判断：对一种学术代偿行为的发生学研究Feigned Judgment: The Genesis of an Academic Compensatory Practice
这篇论文摘要正确、结构完整、查重通过、答辩通过——可你就是感觉不到作者站在文字背后。它不是抄袭，不是能力不足，是第三种东西：判断从未发生，学科话语的语法在替他判断，而他真诚地以为那是自己。制造装判断的，不是体制的松懈，恰恰是体制的严格。
The abstract is correct, the structure complete, every check passed — and nobody is home. Not plagiarism, not incompetence: a third thing. Judgment never occurred; the grammar of the discipline judges in his place, and he sincerely believes it is him.
读全文 →Read → 王德生 著 · 约 2.3 万字by Wang Desheng · ~23k words纠葛与秩序：学术原创性的认知发生学批判Tangle and Order: A Genetic Critique of Academic Originality
导师问“你的研究问题是什么”，他说不出——不是懒，是那团让他睡不着的东西在结构上装不进那一栏。本文给它一个名字：纠葛。原创性不是从清晰问题推导出来的结论，而是从纠葛出发、在与秩序的摩擦中不可预期发生的认知突变；抄袭与平庸论文共享同一个根——都是被净化干净的机器加工出来的合规产品。
Originality does not follow from a clear research question; it erupts, unpredictably, from a pre-problem tangle the intake forms have no field for. Plagiarised and mediocre papers share one root: both come off a line that washed the tangle out at the gate.
读全文 →Read → 王德生 著 · 约 2.5 万字by Wang Desheng · ~25k words从失压到无困：论抄袭的深层病理与认知生产危机的新诊断From Lost Pressure to No Perplexity: Rediagnosing the Deep Pathology of Plagiarism
被质问的抄袭者说：“每篇文献都写得很好，我不知道我该怎么写。”这不是道德推脱，也不是被压垮——是认知起点从未发生。失压者体内残留“这里本应有问题”的残根，无困者连这个都没有。种子枯了要浇水；没种过的地，得从头开荒。
The plagiarist says: every paper I read was already so good, I did not know what was left for me to write. Not an excuse, not collapse under pressure — the starting point never occurred. A withered seed needs water; unsown ground must be broken.
读全文 →Read →「我觉得这是原创的」是自陈，不是判据。判据必须能被别人拿去算，且算出来可能是你不想要的答案——一个不会给出坏消息的原创标准，就不是标准。
"I feel this is original" is self-report, not a criterion. A criterion must be computable by someone else, and must be able to return an answer you did not want — a standard that never brings bad news is not a standard.
认知网络私有化：抄袭作为知识生产结构的适应性截留Enclosing the Cognitive Network: Plagiarism as Adaptive Interception in Knowledge Production
抄袭不是“偷”——因为那些被截留的认知片段，本来也不是任何单一个体的排他性创造。不是本属于个人的东西被偷了，而是本属于网络的东西被截留下来强行归了个人。查重系统真正保护的不是原创性，是“作品”这个分配单位在操作上的可行性。制度不是原谅了抄袭，是生产了抄袭者。
Plagiarism is not theft, because what gets intercepted was never any single person's exclusive creation. The checker protects not originality but the operability of the work as a unit of distribution.
读全文 →Read → 王德生 著 · 约 2.2 万字by Wang Desheng · ~22k words原创抵押权：制度化知识生产中原创定义的自我取消The Originality Mortgage: How Institutionalized Knowledge Production Cancels Its Own Definition of Originality
制度从来没禁止你原创。它只是拿走了裁定权——什么算原创、值多少、够不够续聘——并且在原创还来不及发生时，就把它将要生长的时间征用了。查重软件像验钞机，只认纹理，不认质地。抄袭不是违约者的道德溃疡，是这笔从未有人读过条款的抵押契约，在承受力最薄弱节点上的极端兑现。
The institution never forbids originality. It takes the adjudication rights and expropriates, in advance, the very time in which originality would have grown. Plagiarism is not a moral ulcer but the mortgage contract cashing out at its weakest node.
读全文 →Read →引用最强的对手，而不是最好打的那个Cite Your Strongest Opponent, Not the Easiest One
一篇论文最能暴露作者诚意的地方，是它选择跟谁交手。挑一个稻草人来打，读者立刻知道这篇没底气；正面写出「这条界线很窄，我欠他机制」，读者反而开始相信你。稻草人不是修辞失误，是一种可被检测的心虚。
Nothing exposes a paper faster than whom it picks a fight with. A straw man tells the reader there is no confidence here; admitting that the line is narrow and that you owe your opponent the mechanism is what makes you credible.
选题 · 即将发布Planned · Coming soon谁写的，为什么是一个事实问题，而不是礼节问题Authorship Is a Question of Fact, Not Courtesy
AI 时代让署名问题从礼节退回到事实：一篇文章署谁的名，是关于某个人认知能力的一句陈述，可真可假。给自己写的东西打分、给别人的名字挂自己的产出，都不是失礼，是伪证——而一旦伪证被接受，同一套体系里其余所有的评价就同时失效。
The AI era pushes authorship back from etiquette to fact: a byline is a claim about a person's cognition, and claims can be false. Scoring your own text, or signing your output with another's name, is not rudeness but fabricated evidence — and once accepted, every other assessment in the same system goes void.
选题 · 即将发布Planned · Coming soon