← Principles
第 3 章 · 22 分钟
Five steps, and they must not blur together
Each step needs a different capacity; run them together and they contaminate each other.
概念地图
Five-step loop
Goals / problems / diagnosis / design / execution, and the steps must not blur.
深色为本章概念，浅色为其他章
陪练模式
不想只是读？让它带你走一遍。
导师把这一章拆成小步，每一步先问你，再讲；你用自己的话答，它先认下对的部分再纠偏。用你自己的 AI，对话只存在这台设备上。
Locus01
延伸阅读
PDCA and the quality improvement cycle · mid twentieth century
Writes improvement as plan, do, check, act, arguing against leaving improvement to individual conscientiousness; the cost is that the cycle has no separate step for finding problems, so problems tend to surface only at the checking stage.
Foundational
Dalio · the five steps to personal evolution
Set goals / identify problems and refuse to tolerate them / diagnose to the root / design a plan / do the tasks. He argued against running these together; the cost is that each step needs a different capacity while the book assumes one person can do all five well.
Turn
The harder requirement in the same section · the steps must stay distinct
When setting goals, only set goals; when diagnosing, only diagnose. This, not the list, is the chapter's real mechanism — lists like this come in many versions, while 'do not blur them' is something you can actually execute.
Turn
The OODA loop · military decision-making
Observe, orient, decide, act, with the emphasis on cycle speed rather than purity of steps. It sets up a direct contrast here — speed first or purity first, depending on how fast the environment moves — and the debate block lands on it.
Counterexample
机制02
Splitting one thing into five of different kinds
Most people treat solving a problem as one activity, and so use the same head at the start and at the end. The book cuts it into five, collectively the five steps. The list itself is not novel and comes in many versions; what does the work in the argument is that it supplies five different tests — the goal step asks what you want, the diagnosis step asks why, the design step asks how. Split apart, each step gets a completion standard of its own.
机制One activity, through being cut into five parts with different tests, gives each part its own completion standard.
可迁移性测试
Move it to software: requirements, design, implementation and testing each have their own acceptance criteria, and whoever runs them together discovers halfway through building that the requirement was never settled. The isomorphism breaks in that engineering leaves artefacts at each stage, while a person's five steps happen in their head, where only they can tell whether the steps stayed apart.
机制03
Blurring contaminates in a fixed direction
The harder requirement is not 'go through five steps' but 'do not run them together', and the damage has a fixed direction: think about feasibility while setting goals and the goal gets trimmed back to what you can already do; think about solutions while diagnosing and the diagnosis stops at the layer you happen to know how to fix. The reason the second step is called not tolerating problems is the same: it is only meant to put problems on the table, and the moment 'can this be solved' joins in, the unsolvable ones vanish before they are recorded.
机制Two steps run together, through letting the later step's constraint arrive early, quietly trim what the earlier step produces.
可迁移性测试
Move it to gathering requirements: let the person who will implement them also collect them, and what comes back is roughly what they intended to build. The isomorphism breaks in that a review can be staffed by someone else, while one person runs all five steps alone, with nobody to block the backflow.
Derivation04
Why blurring necessarily lowers the output
What is being reconstructed is the thought behind 'keep the steps apart', and the condition it depends on.
The five steps have different completion tests
A premise the author chose, and the whole basis of the requirement. If the tests were the same, keeping them apart would not matter.
Running two steps at once puts both tests in force at the same time
A premise the author chose. It describes what people actually do: wanting something while simultaneously judging whether it is achievable.
When two tests are in force at once, the stricter one governs
A step added by this course to join the two premises above. The book contains no such sentence; it is marked so the reader knows it is ours.
推导 · 0 / 4
Worked derivation05
A goal-setting meeting that never sets one
Two hours in, no goal has been set, and every proposal was met with 'we can't do that'. Which step blurred in?
Each time something is knocked back, ask which step's test was used.
裂缝06
Cracks in this reading
争议地形07
本书主张
The five steps must be executed separately: each has its own test, and running them together lets later constraints trim what the earlier steps produce.
另一种看法
The OODA line holds that in adversarial or fast-moving settings cycle speed is worth more than purity of steps; stretching one cycle to keep every step clean may mean the answer arrives after the problem has changed.
分歧扎在
The disagreement is rooted in whether the situation holds still for one cycle, not in whether separation has value. Both sides grant that having goals trimmed by feasibility is bad.
什么证据能裁决
What would settle it is the ratio of environmental change to cycle length: handle the same kind of problem with long clean cycles and with short fast ones, and compare final outcomes. The comparison needs 'a cycle' and 'outcome' defined in advance, and can be rechecked by a third party.
The balance falls by rate of change: where the situation is largely stable within a cycle, the evidence for separation is stronger; where it changes faster than a cycle, fast looping wins. This book is about long-run personal and organisational development, mostly the first case, so its register is reasonable — but the condition is never written down, and a reader carrying it into a fast-moving setting gets no warning. What is missing is a ready test comparing rate of change with cycle length.
Falsification08
Turning 'separated beats blurred' into a testable specification
The claim under test: setting goals and assessing feasibility in two separate rounds yields wider goals than doing both at once, without worse plans.
Data source: the same people's meeting records on the same set of problems under both procedures, counting goals proposed per round and the plan finally adopted
Both procedures run on the same people, so individual differences do not enter the comparison.
Sample range: at least 20 problem sets, each done once separated and once blurred, in random order
Randomised order, or whichever went first collects the learning effect.
Time window: assess delivery three months after a plan is adopted
Fixed in advance; wider goals that never land do not count as support.
Decision threshold: the separated round proposes at least 50% more goals, with a three-month delivery rate no lower than the blurred round
Both must hold; more goals that never land means separation just produced talk.
Failure condition: no clear increase in goals proposed, or a clearly lower delivery rate
Either counts as the claim being refuted.
推导 · 0 / 3
接口09
Which slot it hangs on
挂在哪个槽位The conclusion you already hold is probably 'work through things in order, one step at a time' — a judgement about pace. This chapter goes at what is unclear in it: what exactly is being separated.
Does this chapter (a) replace 'one step at a time', or (b) constrain it: what gets separated is not the sequence but the tests — two tests must not be in force at once, even within the same hour?
慢变量Register two slow variables: how often you recently said 'we can't do that' during goal setting, and in how many of those you moved it to the next round instead. Look at the second first.
小结10
本章小结
01The list of five is not novel; what works is that it supplies five different tests.
02The real mechanism is 'do not blur': later constraints arriving early trim what earlier steps produce.
03'Blurred output is narrower' follows from two tests taking an intersection; it is not about attitude.
04This course added a premise here (the stricter test governs), and it does not always hold.
05'The five steps were not done well' is an after-the-fact verdict consistent with any outcome, so this course books it as unadjudicable.
提取练习 · 合上书，先自己答一遍。
?Is 'blurred output is the intersection of separated outputs' (a) an unpacking of 'satisfying both' or (b) an empirical claim observation could overturn?
?Is the OODA objection settled by (a) whether separation has value or (b) whether the situation holds still for one cycle?
Forced choice11
A · Identify the mechanism
Which of these is a mechanism?
二选一
Forced choice12
C · Locate the crack
In which situation does this chapter's reading give a confident and wrong answer?
二选一
读到这里，把它记为已读。