Critique of 02-arc.md (draft v1)

Independent adversarial read by a fresh agent, 2026-09-19, briefed as a series editor making this work for a busy newcomer on a phone. Checked against Dettmer ch. 1 and the existing articles. Lightly condensed.

1. Is the story right? Where's the pull?

Series A opens brilliantly and then spends its mystery. "Sixty people waiting … they are all wrong" is the best hook in the repo, but A1's figure (five stations, output set by the smallest) gives away the answer in chapter one: it's the doctor. After that A has only abstractions to offer. Fix: keep a question open. The doctor's day is half admin — why? — unanswered until the current reality tree (A8) traces it to the countersignature rule, and Maria's "same jam from two ends of the corridor" pays off in the last chapter. Right now "At Ashfield" is buried in A10.

A sags at A5–A7: three consecutive chapters where nothing happens at the clinic (five heads, eight questions, then the least dramatic tree). A newcomer on a phone quits there.

Series B's pull is Maren and it holds until B5–B6, where naming an assumption (B5) and dissolving it (B6) is one idea over two chapters, and B6 depends on a demo that may not exist. If B ends on "it's a prototype", the last feeling is "not real yet". Decide whether the demo exists before committing to B6's shape.

2. Is each chapter one point?

• B1's teaser is a scene, not a claim — nobody can repeat it. Merge with B2 so the claim is "groups have the same conversation again and again".
• A4 is two points (the constraint moves; it's usually a rule). The rule is the one that matters, as the hinge into Part two; "it moves" is one line at the end of A3.
• A6 has eight questions: a checklist, not a point.
• A10 is two trees plus a recap.
• B6 is AI filing + people merging + the map talking back: three points. Cut "talks back" to a sentence.
• A1/A2 overlap: A1's teaser is a definition, A2's ("not where people are busiest") is the repeatable claim. Merge.

3. Does the order build?

Missing, and depended on later:

• What an arrow means. There are two logics and the arc never says so: goal tree = necessity ("in order to have X we must have Y"), current reality tree = sufficiency ("if X then Y"). A6's fire example is really sufficiency (the AND); B4's "missing sub-condition" is necessity. Reshape A6 as "An arrow is a claim you can check": the two readings, then three or four fair questions (clarity, existence, causality, is it enough on its own); the full eight in a footnote.
• Order: A6 teaches rules for arrows before the reader has seen a tree. Show a tiny tree first, even two boxes and an arrow from the clinic.
• The bridge between A's halves. A8 must say the line that makes Part one pay off: the root cause the tree finds is the constraint. Dettmer p. 1: root causes "inevitably include the constraint". Without it, Part one and Part two are two essays.
• The five focusing steps are scattered over A1–A4 but never named. A one-picture recap at the end of Part one is cheap and useful.
• Throughput isn't needed as a term; "patients seen, well" is the clinic's, said plainly in A1.

Not needed: prerequisite and transition trees at full weight. For a newcomer, "check the fix forward, then plan the obstacles" is a sentence in the last chapter.

4. Two series or one; which first?

For one: a single pull. B's ideas (addresses, necessary conditions, the dilemma) are A's concepts, so B read cold has to re-teach them, and a coda gives the method a purpose.

For two: different readers (managers vs the Second Renaissance forum), each finishable in a sitting, and B is a pitch for Reason Commons that a method series shouldn't carry.

Recommendation: two, B first on the index. The people arriving now come from the Second Renaissance; B is shorter, has a person in it, and ends on an ask; A is the go-deeper path. B4 must then teach necessity itself in two sentences rather than lean on the link to A7.

5. Too long?

Yes: sixteen chapters is two evenings. Cut to 8 + 5:

• A: (1) sixty people waiting: the constraint isn't where people are busiest; (2) squeeze before you spend, then it moves; (3) the stubborn constraint is an invisible rule, so five people give five answers; (4) an arrow is a claim you can check; (5) the goal tree; (6) the current reality tree finds Ashfield's rule; (7) the cloud; (8) walk it forwards: test the fix, then plan; the five questions; Maria.
• B: (1) groups have the same conversation again and again; (2) nowhere to put it; (3) give every claim an address; (4) why nobody does the filing; (5) talk first, file later, with the worked example.

6. The five open questions

1. Two series, B first.
2. Keep the clinic: real arithmetic, and "the rule nobody can see" is the movement's problem too. Say that bridge in A8's last line.
3. Eight, not ten. Don't split A6; shrink it to three or four questions.
4. No A11. One closing line in A8 ("keeping trees like this alive is the expensive part — that's the other series").
5. "Thinking in trees" misleads, as Part one has no trees; try "Find the constraint". Keep "Where does the thinking go?".

Also: takeaway 2 overclaims ("you can [agree] by drawing…"); the voice guide wants a calibrated hedge — "you have a much better chance".

Top 5 changes

1. Keep a question open through A: the doctor's admin stays unexplained until the current reality tree finds the rule; end A on Maria.
2. Teach what an arrow means (necessity vs sufficiency, a tiny tree first) before the checking rules; three or four questions, not eight.
3. Cut 16 chapters to 8 + 5.
4. In A6/A8, say the root cause is the constraint; add a five-focusing-steps recap at the end of Part one.
5. Settle B's ending before building: worked AI example exists, or end B on the address/tree chapter. Put B first on the index.