Six Divergence Points · Three Trajectories Each · Eighteen Timelines六个分叉点 · 每点三条轨迹 · 十八条时间线

The history we got is one of many that almost happened. 我们所拥有的历史是几乎发生过的众多历史之一。

Synthetic History generates alternate timelines from six well-studied divergence points — Tang continuance, no Industrial Revolution in Britain, no Black Death, Cold War turning hot in 1962, no internet by 2000, no GPT moment in 2022 — and runs three plausible trajectories from each. The output is templated, not novelistic; the value is the honesty about what counterfactual reasoning can and cannot tell us. 合成历史从六个经过良好研究的分叉点——唐朝延续、英国无工业革命、无黑死病、1962 年冷战变热、2000 年仍无互联网、2022 年无 GPT 时刻——生成替代时间线，并从每点跑三条可信轨迹。输出是模板化的，不是小说式的；价值是对反事实推理能与不能告诉我们什么的诚实。

six divergence pointsthree trajectories eachhonest counterfactualsEN · 中文
S1 · Timeline Generator

Pick a divergence point. Pick a trajectory. Read the alternate.挑一个分叉点。挑一条轨迹。读那条替代。

6 points × 3 trajectories = 18 templated alternate timelines. Each output names the divergence, the immediate downstream effect, the 50-year arc, and the meta-honest read on what this counterfactual does and does not actually tell us.6 个点 × 3 条轨迹 = 18 条模板化替代时间线。每个输出命名分叉、即时下游效应、50 年弧线，与"这个反事实实际能与不能告诉我们什么"的元诚实读数。

S2 · Reader

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

S3 · Guardrails

What this generator refuses to pretend.本生成器拒绝伪装的事。

Five things Synthetic History will not do.合成历史不会做的五件事。

  • No prediction wearing alt-history clothes. Counterfactual generation is not forecasting. People who treat the outputs as predictions of what would have happened are using them wrong; the outputs are stress-tests of how we think about contingency.不做穿替代历史外衣的预测。反事实生成不是预测。把输出当作"会发生什么"的预测的人在错用；输出是关于我们如何思考偶然性的压力测试。
  • No ethnic flattering. The generator refuses to produce timelines that exist mainly to flatter a specific ethnic, national, or political identity. Honest alt-history is uncomfortable for everyone.不做民族奉承。生成器拒绝产出主要存在以奉承特定民族、国族或政治身份的时间线。诚实的替代历史让所有人都不舒服。
  • No "great man" determinism. The generator does not output timelines centred on individual leaders changing all of history. The structural-cause weighting is built in; if you remove a single individual the world rarely diverges as much as the genre suggests.不做"伟人"决定论。生成器不输出以个别领导人改变全部历史为中心的时间线。结构性原因加权已内建；移除单一个人，世界很少像类型暗示的那样大幅分叉。
  • No moral arbitrage. The generator does not assert any timeline is morally superior to ours. Counterfactual ethics is famously hard; we describe trajectories without ranking them.不做道德套利。生成器不断言任何时间线在道德上优于我们的。反事实伦理出名地难；我们描述轨迹而不排序。
  • No infinite-AI-generation slop. 18 templated timelines + 5 canned probes is better discipline than infinite plausible-sounding LLM alt-history.不做无限 AI 生成的污水。18 条模板化时间线 + 5 个预设探针，比无穷个貌似合理的 LLM 替代历史是更好的纪律。