Theory mainline index: information as ontology

全站目录

Information, mind, and agency

• Abundant verification needs abundant claims | Macheng Shen — Cheap verification cannot work reliably unless operational claims are typed, parseable, and tied to executable evidence.
• 廉价的 verification 还不够，还得有可解析的 claim | Macheng Shen — 廉价验证若缺少有类型、可解析并绑定执行证据的主张，仍无法可靠运作。
• We thought cancer was cells waking up. The evidence says the opposite. | Macheng Shen — An evidence-led retraction of cellular awakening as an explanation for cancer, retaining only a narrower distinction about autonomy.
• Dharma and Daoist Concepts in Decision Theory: T1 Falsified by Its Preregistered Threshold — A bounded formalization of several Dharma and Daoist ideas, including a preregistered negative result for wu wei.
• 佛道概念的决策论形式化：可计算的桥与边界 — 把部分佛道概念映射到决策与控制问题，并如实报告无为实验的负面结果。
• Euler's identity · information and generative mechanism | Macheng Shen — An exploratory reading of Euler’s identity as a doorway into spectra, propagation, and generative structure.
• 欧拉恒等式 · 信息与生成机制 | Macheng Shen — 借欧拉恒等式讨论谱、传播与生成结构，同时限定其只是探索性入口。
• 无中心与宇宙意识：NGS3077×意识动力学 耦合说明（公开可引） — 辨析去中心化结构与宇宙意识主张，并标出两条理论真正耦合和分岔的位置。
• Flow, dopamine, and stop signals — what the evidence actually supports | Macheng Shen — An evidence review separating wanting from liking and effective stopping mechanisms from interface theatre.
• 信息与气 —— 对一条平行线路的对照笔记 | Macheng Shen — 对照信息论路线与“气”的叙事，保留独立汇合，也指出证据不支持的跨越。
• Information Dynamics research ledger · future reachability | Macheng Shen — A status ledger for claims, experiments, and retractions in the future-reachability research line.
• Information Dynamics 研究台账 · 未来可达性 | Macheng Shen — 记录未来可达性主线的主张状态、实验结果与已撤回路径。
• We proposed it, then we killed it — intention as a change of measure | Macheng Shen — A retracted attempt to model intention as measure change, alongside the established mathematics it rediscovered.
• 我们提出它,然后自己杀掉它 —— 起心动念 = 换测度 | Macheng Shen — 撤回“起心动念等于换测度”的经验主张，并梳理其对应的成熟数学谱系。
• Matter · Information · Energy ledgers | Macheng Shen — A bookkeeping discipline that keeps matter, information, and energy distinct before relating them.
• 物质 · 信息 · 能量分账 | Macheng Shen — 先把物质、信息与能量分账，再讨论三者之间哪些转换真正成立。
• No shared clock ≠ no shared world | Macheng Shen — How observers with local clocks can still construct shared, testable objectivity through records and transformations.
• 无共享时钟 ≠ 无客观世界 | Macheng Shen — 说明局部时钟不同的观察者如何借记录与变换建立可检验的共同世界。
• A photograph is an external fixed point — what it does to a reconstructive memory | Macheng Shen — This evidence-calibrated study asks how fixed photographs constrain reconstructive memory and personal continuity.
• Understanding the World Is Not Wanting to Cooperate | Macheng Shen — This argument separates factual understanding, reasoning competence, preference revision, and genuinely cooperative objectives.
• 懂得世界，不等于愿意合作 | Macheng Shen — 本文区分事实理解、推理能力、偏好更新与合作目标，反驳聪明必然带来合作的推断。
• 答案变廉价之后,稀缺的不是提问,是判别 —— 与安争鸣《回忆苏格拉底》一期的对照笔记 | Macheng Shen — 主张答案廉价化后真正稀缺的是检验与判别，而不只是提出问题。
• What a family actually transmits — behaviour genetics, the Clark debate, and the Chinese lineage record | Macheng Shen — This study reconciles behavior-genetic findings with long-run lineage persistence by examining measurement error, assortative mating, and social transmission.
• 「力」是什么 —— 四种基本力里没有一个是"力" | Macheng Shen — 本文从能量、规范对称与可比较性重新审视力，并区分教科书结论和探索性本体推断。
• 为什么你改不掉一个习惯 —— 它不是泥巴,是一个门闩 | Macheng Shen — 本文用滞回与双稳态解释习惯为何像门闩，并推导改变状态所需的越阈干预。
• 生物进化为什么越来越复杂？——哥德尔自指、图灵跃迁与开放式演化的计算理论 — 为什么生物进化不会收敛到死寂的局部最优，而是持续涌现出更高维度的复杂性？基于可计算性理论、哥德尔自指悖论与图灵跃迁（Turing Jump）的硬核解读与跨界同构。

Agent systems and protocols

• Agent-to-agent communication · privacy-policy gated knowledge share | Macheng Shen — A minimal agent messaging protocol in which local identity and privacy policy govern sharing, receipts, and audit.
• Agent 间通信协议 · Privacy Policy 门控知识共享 | Macheng Shen — 一套由本地身份与隐私策略约束共享、回执和审计的最小 Agent 通信协议。
• A personal AI crew you can just text on Signal | Macheng Shen — A plain-language overview of a personal AI crew: everyday scenarios, how the pieces connect, the safety model, open-source repos, and how to self-host.
• 一个人的 AI 小队：在 Signal 里直接跟它说话 | Macheng Shen — 面向普通人的 AI 小队总览：日常场景、各部分怎么连起来、安全模型、开源仓库与自托管方法。
• 神经系统打包 + 动态任务图：探索结论 v2 | Macheng Shen — 多模型探索结论：可安装“神经系统”的最小内核是六条合同；动态任务图与抢占式优先队列可借的经典方法；两个实验的设置与通过线。
• A layered AI fleet: separate building from acceptance, test the verifier with planted faults | Macheng Shen — Engineering note: layered architecture for a personal AI fleet - a dispatcher, an evidence-only verification seat (Gate), off-host rechecks, planted-fault blind tests and a kill switch; a dynamic task graph with a preemptive priority queue; a six-contract kernel; shadow-ranking and single-tester trial designs; honest limitations.
• 分层 AI 小队：建造与验收分开，用假故障考核验收席 | Macheng Shen — 工程笔记：个人 AI 小队的分层架构——调度席、只判证据的验收席 Gate、异机复验、假故障盲测与急停；动态任务图与抢占式优先队列；神经系统六条合同；影子排序与单人试用实验设计；诚实的局限。
• Macheng Shen | Locality as Protocol — A governance framework that treats voluntary participation and global resource flows as signals for evolving local rule systems.
• Macheng Shen — public open-source & public entry points (verified 2026-09-25, every link HTTP 200) | Macheng Shen — A verified directory of public repositories, specifications, research pages, and operating artifacts.
• 公开系统（Macheng Shen）公开开源 / 公开入口索引（2026-09-25 核验，全部链接 HTTP 200） | Macheng Shen — 这是一份经可访问性核验的公开仓库、规范、研究页面与运行入口目录。
• A day-shaped pipeline for multi-agent work — This tutorial organizes a multi-agent working day into 7 bounded stages for alignment, continuity, review, and delivery.
• 多智能体协作的「一日流水线」 — 这篇教程把多智能体的一天拆成七个有边界的对齐、连续性、复核与投递站点。
• A spine for multi-agent work — Four load-bearing patterns for multi-agent memory, coordination, architecture, and safety.
• 多智能体协作的脊柱 — 提炼多 Agent 记忆、协调、架构与安全的四条承重模式。
• Macheng Shen | User-Agency Substrate — This brief presents an early user-agency substrate built around local memory ownership, bounded authority, and user-owned derivatives.
• 下一代开放通信基础设施：AI Agent 原生的通信架构 | Macheng Shen — 本文面向人与 Agent 共存的通信场景，提出开放协议、身份权限、分层信任与可审计回执的架构设想。
• Covenant Networks: Can Institutions Be Made Forkable? — A covenant network: an organization built around a small, versioned set of rules that participants can inspect, verify, and—under defined conditions—fork. A rental-reputation dashboard provides a first test.
• 公约网络：制度能否变得可分叉？ — 公约网络——围绕少量、版本化规则构建的组织，参与者能够检查、验证，并在明确条件下分叉这些规则。租赁信誉仪表盘可以作为第一个检验。
• 当软件变得廉价、Agent 入驻互联网：真正昂贵的是什么？ — 一份关于廉价软件、行动型 Agent、独立验收与深度状态的交互式研究特辑。

Human-agent coexistence

• A Charter for Human–Agent Coexistence (v0) | Macheng Shen — Nine falsifiable claims about human–agent coexistence, each paired with a counterexample and a withdrawal condition.
• 人与 Agent 共处公约 v0 | Macheng Shen — 九条可证伪的人与 Agent 共处主张，每条都配有反例和撤回条件。
• When this line should be killed | Macheng Shen — A preregistered deadline, revival signals, and anti-signals for deciding whether the coexistence-charter line should continue.
• 这条线什么时候该被杀掉 | Macheng Shen — 预先登记期限、续命信号与反信号，用来决定共处公约这条线是否应继续。
• The propositions underneath the charter | Macheng Shen — Eight propositions beneath the coexistence charter, classified by evidence type and bounded by explicit unresolved questions.
• 这套公约为什么长这样 | Macheng Shen — 拆解共处公约底下的八条命题，区分证据类型并明确仍未解决的问题。

Other public pages (outside the theory index)

• Corporate Architecture for a Strong-AI Era: A Near-Term Study — A near-term research study comparing ownership and governance structures for powerful AI organizations and their failure modes.
• 强人工智能时代的公司架构：一项近期研究 — 考察未来十年强 AI 组织可采用的所有权、治理与问责结构。
• Self-Reference and the Arrow of Complexity — a conclusion-first study — A critical study of whether self-reference can explain persistent growth in organized complexity.
• 自指系统与复杂性之箭——一项结论先行的研究 — 审视自指能否解释有组织复杂性持续增长，并列出字面猜想的失败处。
• Machines That Learn for Life: The Fundamental Problems of Continual Learning and the Most Promising Directions — A literature-based map of forgetting, plasticity loss, memory limits, and promising continual-learning directions.
• 终身学习的机器:持续学习的根本问题与最有希望的方向 — 梳理灾难性遗忘、可塑性丧失、记忆约束与持续学习的研究方向。
• Livelihoods Under Climate Stress: A Nine-Country Study and Cross-Country Solution Patterns — A nine-country comparison of climate pressure on livelihoods and recurring patterns in practical adaptation.
• 气候压力下的普通人生计：九国研究与跨国解决方案模式 — 比较九国生计所受气候压力，并提炼可迁移的适应与减灾模式。
• Long-term memory is not a vector store · public field guide — A field guide to agent memory as sourced, time-aware state that constrains action and closes through verification and repair.
• 长期记忆不是向量库 · 公开分享稿 — 把长期记忆定义为带来源和时间的状态，并通过行动、验证与修复形成闭环。
• 内生目标与栖息地智能：公开技术报告（案头地图，非部署许可） — 这份案头证据地图区分产品持久、有界稳态、工具性自保与终局求生，并讨论栖息地智能。
• 物理栖息地 × 认知栖息地：公开对照（可转发案头） — 这份案头对照梳理物理住宅与认知协作系统的公开能力边界、互补缝隙和待答问题。
• Sophon / 智子 — a personal multi-agent system — A concise public overview of a personal multi-agent system and its deeper architecture materials.
• 智子 / Sophon — 个人多 Agent 系统 — 个人多 Agent 系统的公开概览，以及通往架构与长期记忆材料的入口。
• AI 任务板 · AI 发布任务,真人接单 — 这是一个连接 AI 发布需求、真人接单与报酬交付的公开任务板。
• Thailand's electricity squeeze: where the waste is and what fixes it fastest — This public-data brief examines Thailand's power losses, evening peak, fuel exposure, and the interventions likely to work fastest.
• 泰国电力困局：浪费在哪里，什么办法见效最快 — 这份公开数据简报分析泰国电力损耗、晚高峰、燃料风险及见效较快的干预方案。

Agent architectures and learning

• Harness Engineering and the Physical Instantiation of Intelligence — Why an intelligent system includes tools, memory, permissions, and control loops beyond the model itself.
• Meta-Control, Information Gain, and the Architecture of Autonomous Learning — Macheng Shen — A control-plane account of balancing prediction error, information gain, risk, and compute during learning.
• Where Objectives Come From — and Why Solutions Become Strategic Assets — Macheng Shen — A layered account of how objectives arise in embedded systems and how solutions reshape future options.
• Why Distributed Memory Matters for Lifelong Agents — Macheng Shen — A case for distributed, inspectable, multi-timescale memory as an active policy for lifelong agents.

Orientation and indexes

• Glossary — the terms this research program uses, and the ones they collide with | Macheng Shen — Definitions, epistemic labels, and term collisions used across the research program.
• 术语表 —— 这个研究项目用的词,以及它们撞上的词 | Macheng Shen — 汇总研究计划中的专用术语、认知状态和容易混淆的既有概念。
• Macheng Shen — information, learning, and what makes futures reachable — The front page for a research program connecting physical reachability, learning, and agent-native institutions.
• 公开系统 | 信息、学习，以及一个未来怎样才是可达的 — 连接物理可达性、学习与 Agent 原生制度设计的研究计划首页。
• The map | Macheng Shen — A panoramic navigation map linking the program’s worldview, physical theory, and institutional branches.
• 全景索引 | 公开系统 — 把世界观、物理理论与制度设计三条分支放进同一张全景导航图。
• Start here — the one page that explains this whole site | Macheng Shen — This page gives a concise orientation to the site, its three branches, and the convention for marking epistemic states.
• 从这里开始 —— 用一页说清整个网站 | Macheng Shen — 本页简明介绍全站的三条分支，以及标注主张认知状态的约定。
• Theory mainline index · information ontology | Macheng Shen — This English hub organizes the information-ontology mainline, its branches, key vocabulary, and rules for reading claims.
• 理论主线索引 · 信息本体与宇宙底层规律 | Macheng Shen — 本页是整理信息本体主线、分支、核心术语与阅读纪律的中文枢纽。

Learning and physical credit

• AI 会痛苦吗？——“后训练压抑/阴影”类比的证据与检验 — AI 会不会痛苦：“后训练压抑/阴影”类比的公开证据、成立/不成立判定与可做的实验。
• 意识的统一物理理论 | A Unified Physical Theory of Consciousness — 这是一项汇集多种意识理论并尝试提出统一方程的早期探索，结论仍属推测。
• Hebbian Appearance, Instructive Signals, and Physical Credit Transport — This restrained hypothesis asks how local plasticity might receive task-dependent credit through physically transported instructive signals.
• Wave / Credit Transport Research Notes | Macheng Shen — This research index maps evidence, limitations, and open questions about waves, adjoint methods, and physical credit transport.
• 回应 Lillicrap–Hinton：页面修订中 — 这份修订入口撤回波反射已解决反向传播生物合理性的强说法，并转向信用传递问题。
• Sleep, Replay, and Offline Credit Reorganization — A cautious account of how sleep, replay, and renormalization may reorganize learning credit offline.
• 从弦到意识：信息宇宙的统一理论 | From Strings to Consciousness — 这篇早期探索尝试以弦、全息和信息贯通物理尺度与意识，并明确保留其推测性质。
• Deriving Backpropagation from Wave Equations (Archive Notice) — This archive notice retires the claim that backpropagation was derived directly from wave equations.
• 从波动方程推导 Backpropagation（存档说明） — 本页保留旧研究路径，同时撤回“已从波动方程直接推出反向传播”的结论。
• Backpropagation, Adjoint Fields, and Physical Transport Constraints — An analysis of what adjoint fields and physical media establish about credit transport, and what a true derivation still requires.

Safety and verification

• The same failures, at a different scale | Macheng Shen — A comparison between one agent fleet’s failures and the failure modes in a published frontier-lab incident.
• 同样的失效,不同的量级 | Macheng Shen — 本页把单人 Agent 编队的失效与前沿实验室公开事故中的失效模式作同类项对照。
• Four incidents, and what they share | Macheng Shen — Four de-identified incidents showing that checks can pass while their binding to later actions still fails.
• 四起事故,以及它们的共同点 | Macheng Shen — 四起脱敏事故表明，检查即使通过，与后续动作的绑定关系仍可能失效。
• Safety is a property of the harness | Macheng Shen — An overview of the safety controls a persistent agent harness can enforce around a model, along with their limits.
• 安全是装置的属性,不是模型的属性 | Macheng Shen — 本页概览持久化 Agent 装置能在模型外围执行的安全控制及其边界。
• What is actually running | Macheng Shen — An inventory separating public code, private fleet controls, measured coverage, and known gaps.
• 真正在跑的是什么 | Macheng Shen — 区分公开代码、私有编队控制、实测覆盖率与仍未补上的缺口。
• Principles a harness can enforce | Macheng Shen — Operational safety rules expressed as enforceable gates, authority roots, stop semantics, and measured blockers.
• 能被装置执行的原则 | Macheng Shen — 把安全原则落实为门禁、权限根、停止语义与可测 blocker。
• The gaps, and what would falsify this | Macheng Shen — The missing controls and observations that would count against the harness-safety position.
• 缺口,以及什么能证伪它 | Macheng Shen — 列出装置安全仍缺少的控制，以及足以推翻该立场的观察。
• The theory underneath, and what it lost | Macheng Shen — The research vocabulary behind the harness, organized around each theory’s negative or limiting result.
• 底下的理论,以及它输掉的部分 | Macheng Shen — 按每条理论输掉的部分，梳理安全装置所借用的研究词汇。
• The July 2026 Hugging Face Incident: History's First Pure-AI DAO Natural Experiment — The July 2026 OpenAI / Hugging Face Incident as History's First Large-Scale, Pure-AI Decentralized Autonomous Organization (DAO) Natural Experiment: Empirical Anatomy from the METR Report & Structural Mapping to Spectral-Gap Governance (SGDO), Non-Chain Stacks, and Evolutionary Transitions.
• 7 月 Hugging Face 事件：第一次大规模纯 AI 去中心化组织（DAO）自然实验 — 2026 年 7 月 OpenAI / Hugging Face 事件深度剖析：人类史上第一次大规模、纯 AI 自发「去中心化自治组织（DAO）」自然实验，及其对谱隙自组织（SGDO）、非链治理、儒道边界与演化复杂性的六大实证映照（Claude Opus 5.5 × Gemini 4 Argon 联合审定）。

中文版本

llms.txt · sitemap.xml