Lab · 2026.08.25
实验需要反馈回路Experiments need feedback loops
关联项目：Lab · WimDataRelated projects: Lab · WimData
实验的目标不是让项目列表变长，而是减少不确定性。一个小原型只有在结果能改变下一步判断时才有价值。
先写清问题
在选择工具和实现方式前，先说明要验证的假设、能够观察到的结果以及停止条件。这样可以避免把“已经做了很多”误当成“已经学到很多”。
记录真实结果
不使用虚构的活跃项目数、完成率或进度条。实验状态用自然语言说明，并保留能支持判断的实际证据。
让结果回流
有效结果进入 WimChat、WimData 或知识花园。无效结果也应留下原因，使下一次实验从已知边界继续，而不是重复开始。
The goal of an experiment is not to make a project list longer. It is to reduce uncertainty. A small prototype matters only when its result can change the next decision.
Write the question first
Before choosing tools or implementation, define the hypothesis, observable result and stopping condition. This prevents confusing “a lot was built” with “a lot was learned.”
Record real results
Do not use invented active project counts, completion rates or progress bars. Describe status in plain language and keep actual evidence that supports the decision.
Return the result
Useful results enter WimChat, WimData or the knowledge garden. Unsuccessful results keep their reasoning too, so the next experiment continues from known boundaries instead of restarting.