This course does not retell the book's entries. It takes the practice apart as a device: upgrade one judgement into a reusable rule, then treat what you are responsible for as goals → machine → outcomes, reading the gap between outcome and goal as the machine's instrument reading. Every step is marked as either a premise the author chose, something that follows from his own definitions, or an empirical claim that observation could overturn. The book almost never states failure conditions, so each chapter's falsification block sets out how the claim would have to be tested — not a test the book already ran. The last chapter audits the first six. It is an independent re-explanation based on the 2011 PDF Bridgewater released free; it teaches the framework rather than reproducing the text, and every example is our own.
结构交底
本课交付的结构
A reusable rule, through replacing 'what do I do this time' with 'what do I do with this kind', makes one piece of judgement pay for itself once.
It carries to any situation that recurs: ask whether this call can be written as a rule that works next time. It fails where nothing recurs — the rule never earns back its cost.
核心承诺
01Bad outcome, fix the machine Outcomes are what the machine puts out; change the outcome without changing the machine and the same outcome comes back.
02Diagnose down to the root Proximate causes are verbs, root causes are adjectives; leave the root in place and the problem returns.
03Opinions are not equal Weight opinions by track record and reasoning, not by headcount or rank — but first ask who assigns the weights.
掌握路径
L1Restate and bound
学完能
Say how a principle differs from an answer to just this one case
Say how goals, machine and outcomes give you a reading