Before you name it,
Falsify it.

在命名之前，
先证伪。

A systematic falsification methodology that kills bad ideas before they cost you time, code, or money.

一套系统化的证伪方法论——在浪费时间、代码和金钱之前，杀死坏点子。

| The Problem问题

Every new idea feels promising at birth. You name it, build a model, run tests — only to discover months later that you were measuring noise, or rediscovering something already known under a different label.

每个新想法在诞生时都充满希望。你给它命名、建模型、跑测试——几个月后才发现测的是噪音，或者只是给已知的东西换了个标签。

This is a validation-order problem: we systematically test for confirmation before testing for falsification.

这是验证顺序的问题：我们总是先验证它是对的，而不是先问它在什么条件下是错的。

| Stage Zero阶段零

Stage Zero processes any fuzzy concept — a market intuition, a theoretical hunch, a pattern you keep seeing — through nine procedural gates before a single line of code is written.

阶段零将任何模糊概念——一个市场直觉、一个理论猜想、一个反复出现的模式——经过九道程序化关口过滤，才允许进入一行代码。

The Equivalence Threshold

等价性门槛

If your idea correlates r > 0.9 with any existing benchmark after structural isolation, it carries zero new information. The process terminates.

如果核心机制剥离后，你的想法与任何已知基准的相关性 r > 0.9，则携带零增量信息。流程终止。

Falsification-First

证伪优先

Every surviving concept must survive a counterfactual attack: under what conditions does this idea fail completely?

每个存活的概念必须通过反事实攻击：这个想法在什么条件下会彻底失效？

| Case Studies案例

The methodology was forged in the failure of three ambitious projects. Each survived months of conventional validation before Stage Zero revealed their structural emptiness.

这套方法论诞生于三个失败项目。每个都通过了数月的传统验证，直到阶段零揭穿了它们的结构性空洞。

D(S) — Structural Distortion
D(S) — 结构畸变指标
r ≈ 0.964
EQUIVALENT
等价
Market structure index indistinguishable from PCA first component. Zero crisis detection.
市场结构指标与PCA第一主成分无法区分。零危机探测。
ALED — Semantic Entropy
ALED — 语义熵框架
r ≈ 0.964
EQUIVALENT
等价
Five-dimensional concept tracking mapped to known statistical aggregates.
五维概念追踪系统映射到已知统计量。
SMRE — Risk Field Model
SMRE — 风险场模型
N/A
DEMOLISHED
证伪
Counterfactual attack revealed no regime where the model provides unique signal beyond existing volatility estimators.
反事实攻击发现：任何市场状态下，模型均无超越波动率已有估计器的独立信号。

| The Pipeline管线

  1. Discovery — Extract raw phenomenon without naming or theorizing
  2. Multi-perspective projection — Map through ≥3 distinct frameworks
  3. Core extraction — Identify shared mechanism across frameworks
  4. Difference matrix — What does this see that each misses?
  5. Equivalence coarse screen — Discard if r > 0.9 with known constructs
  6. Structural isolation — Strip away all accidental features
  7. Reconstructed spec — Rebuild from core mechanism only
  8. Equivalence audit — Second-pass, stricter check
  9. Counterfactual attack — Prove it can fail, or discard
  1. 发现 — 提取原始现象，不命名、不理论化
  2. 多视角投影 — 通过 ≥3 个不同理论框架映射
  3. 核心提取 — 识别所有框架共同指向的机制
  4. 差异矩阵 — 它看到什么，每个框架忽略了什么
  5. 等价粗筛 — 与已知构造 r > 0.9 则丢弃
  6. 结构剥离 — 移除所有偶然特征
  7. 重建规格 — 仅从核心机制重建
  8. 等价审计 — 第二轮、更严格的等价检查
  9. 反事实攻击 — 证明它可以失效，否则丢弃

Survivors proceed to domain-specific translation — financial, biological, whichever field the concept belongs to.

存活者进入领域特定翻译阶段——金融、生物、网络，取决于概念所属领域。

| Access获取

Stage Zero is available as a structured report service. Each validation run produces a full protocol covering every step of the pipeline.

阶段零以结构化报告服务形式提供。每次验证运行产出完整协议文档，涵盖管线每一步。

Report delivery is gated by structural review to ensure methodological integrity.

报告交付需通过结构性审核，确保方法论严谨性。