The personalities you wear, before you remember wearing them. 那些你已忘记自己正在佩戴的人格。

Six contexts. Six personalities. One self underneath.六种语境，六重人格，下面是同一个自我。

A "personality" isn't a single object. It's a context-conditioned policy network — a different one fires depending on who's watching, what's at stake, and what the room rewards. Below: the six most studied contexts."人格"不是一个单一物体，而是一组随语境切换的策略网络——不同的网络在不同的观察者、利害与房间奖励下被激活。以下是研究最多的六种语境。

Twelve traits × six contexts. The trait that's stable across the row is the closest thing to "you."十二种特质 × 六种语境。整行都稳定的那一项，是最接近"你"的东西。

Each cell is the AI estimate of how strongly a trait expresses in a given context (0 = absent, 9 = dominant). Look for stripes (context-stable = identity-core), gradients (context-shifted = adaptive), and inversions (context-reversed = suppression sites).每个单元格是该特质在给定语境下表达强度的 AI 估计（0=缺席，9=主导）。横向条纹（语境稳定）= 身份核心；横向渐变（语境位移）= 适应；横向反转（语境逆转）= 压抑点。

Every transition has a cost. The expensive ones are where you fracture.每次切换都有代价。代价高的那条边，是你裂开的地方。

A radial graph of context-to-context transitions. Edge thickness = transition frequency. Edge color = cognitive cost (red = high, gold = medium, green = low). The triangle work ↔ romantic ↔ family is where most adults pay the highest tax.语境间切换的径向图。边的粗细 = 切换频率。边的颜色 = 认知代价（红=高，金=中，绿=低）。职场 ↔ 亲密 ↔ 家庭三角是大多数成年人交税最重的地方。

Eight desire archetypes. The mask is what shows; the desire is what makes the mask necessary.八种欲望原型。面具是显露之物，欲望是让面具变得必要之物。

Surface behavior is what others see. Latent indicator is what the AI watches for. Don't read this as a typology — read it as a checklist of vectors that pull on every mask you wear.表层行为是他人所见，潜在指标是 AI 所追踪。不要把它读作类型学，而要读作牵引你每一重面具的向量清单。

Seven defenses. Each one buys you something — and charges you something.七种防御。每一种都为你买来某物——也向你收取某物。

Defenses aren't pathologies. They're load-bearing. The question isn't whether you use them — you do — but which ones, when they trigger, and whether the price stays affordable.防御不是病理，它们是承重的。问题不在于你是否使用它们——你确实在用——而在于使用哪一种、何时触发，以及代价是否仍然可负担。

Suppression is cheap per-event. It's the integral that breaks you.压抑在单次事件上是便宜的。让你崩溃的是它的积分。

Move the sliders. Watch the accumulation curve and the predicted breach window. The model is intentionally simple — three coefficients in a saturating exponential — because the point isn't precision; it's the shape of the cost.移动滑块。观察累积曲线与预测的破裂窗口。模型刻意保持简单——三个系数构成的饱和指数——因为重点不是精确度，而是代价的形状。

Five canned probes plus a free-text fallback. Five well-formed answers beat infinite plausible-sounding text.五个预设探针 + 自由文本回退。五个成型的回答胜过无穷个貌似合理的回答。

Each probe analyses a multi-mask interaction. The free-text box runs a heuristic over your input — not a frontier model. Honesty: this is a tool for surfacing patterns, not a clinical instrument and not a substitute for a therapist who actually knows you.每个探针都分析一种多面具互动。自由文本框对输入运行启发式——不是前沿模型。坦诚：这是一个用于浮现模式的工具，不是临床仪器，也不能替代真正了解你的治疗师。

What this engine refuses to pretend.本引擎拒绝伪装的事。