Six Node Types · One Distributed Network · Honest Map六种节点 · 一张分布式网络 · 诚实地图

Research is no longer a building. It is a network — and most of it is invisible. 研究不再是一座建筑。它是一张网络——其大部分不可见。

Intelligence Network maps the six node types in the emerging distributed AI-human research civilization — labs, individuals, agentic clusters, archives, simulators, governance councils — and the actual flows of insight between them. The flows are not what the genre advertises; the network is real and most of it is not at the institutions you know. "智能网络"映射涌现中的分布式 AI—人类研究文明里的六种节点——实验室、个人、代理式集群、档案、模拟器、治理理事会——以及它们之间真实的洞见流动。流动不是类型所宣传的；网络真实，其大部分不在你所知的机构里。

six node typesflow maphonest visibilityEN · 中文
N1 · Six Node Types

Six node types. The flows between them have changed faster than the institutions.六种节点。它们之间的流动比制度变化得更快。

A node in this network is anything that ingests information, produces a transformation, and emits a reusable output. Not all are institutions; not all are human. The 25-year academic-publishing equilibrium is breaking; the new equilibrium has not yet stabilised.本网络里的节点是任何摄取信息、产生变换、输出可重用产物的东西。不全都是机构；不全都是人类。25 年的学术出版均衡正在破碎；新均衡尚未稳定。

N2 · Reader

Five canned probes plus a free-text fallback.五个预设探针 + 自由文本回退。

N3 · Guardrails

What this map refuses to pretend.本地图拒绝伪装的事。

Five things Intelligence Network will not do.智能网络不会做的五件事。

  • No utopian decentralisation. Distributed does not mean equal; the new network has its own concentration patterns, which the genre often omits because they don\'t flatter the decentralisation narrative.不做乌托邦去中心化。分布式不意味平等；新网络有自己的集中模式，类型常省略它们，因为它们不奉承去中心化叙事。
  • No "academia is dead" framing. Universities remain load-bearing for many slow research programmes; the network adds nodes around them rather than replacing them.不做"学术界已死"框架。大学对许多慢研究项目仍承重；网络在它们周围增加节点而非替代之。
  • No agentic-cluster hype. AI agentic research is a real and growing node type, but its current outputs are far weaker than the demos suggest. The map weights it according to what it actually produces.不做代理式集群炒作。AI 代理研究是真实且增长中的节点类型，但其当前输出远弱于演示所暗示。地图按其实际产出加权。
  • No flat egalitarianism. The map shows actual influence, not ideal influence. Some nodes punch above their weight; some punch below; honest mapping is uncomfortable for both.不做扁平平等主义。地图展示实际影响，不展示理想影响。某些节点超出其权重；某些低于；诚实映射让两者都不舒服。
  • No infinite-AI-generation slop. Six node types + five probes is better discipline than infinite plausible-sounding LLM research-meta prose.不做无限 AI 生成的污水。六种节点 + 五个探针，比无穷个貌似合理的 LLM 研究元散文是更好的纪律。