Arc — the illustrated introduction (v2, for Rufus)

Beads reasoncommons-bnl.7. Revises 02-arc.md after the independent critique. Adopted: cut 16 chapters to 8 + 5, keep a question open through Series A, teach what an arrow means before the rules for checking one, join A's two halves on "the root cause is the constraint", B first, and B ending on a worked example. Differences from the critique are marked.

The reader

Someone who runs or cares about something complex — a team, an organization, a movement — and has felt that effort isn't adding up. Smart, busy, not a management-science reader, often on a phone. Arrives from the Second Renaissance forum, a talk, or a link. Has never heard of the Theory of Constraints or the Logical Thinking Process.

What they should be able to say afterwards

1. Most effort in a system doesn't move the goal; one constraint does, and it is rarely where people are busiest, and often a rule.
2. Arguing rarely settles what the constraint is. Drawing the cause and effect, and checking each arrow with a few fair questions, gives you a much better chance.
3. Five questions any real change has to answer — what should be happening, why isn't it, why hasn't it been fixed, will the fix work, what do we do first — each answered by a differently shaped tree.
4. A group's thinking only accumulates if each claim has an address, and AI now makes it affordable to give it one without making contributors do the filing.

B carries 4; A carries 1–3.

Shape

Two short series, each readable in one sitting (~15 minutes), cross-linked. Series 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. B teaches the little method it needs (necessity, in two sentences) rather than depending on A.

Series B — Where does the thinking go? (5 chapters)

Running character: Maren (composite, said so). Every tree shown is the Second Renaissance's real one.

1. The same conversation, again and again — Groups keep having the same good conversation, because what one person learns compounds but what the group agrees keeps resetting to zero. · Maren's post arrives, gathers nine replies, greys into the thread (one looping animation); the research circle reaching the same point three times; the person's staircase vs the group's peaks. The movement is introduced after Maren, in two plain sentences.
2. Nowhere to put it — Forums file by date and papers by paragraph, so a single claim has no address, and what you can't point at you can't improve. · "A claim on page forty has no address." The newcomer's specific disagreement is the thing we're least able to receive.
3. Give every claim an address — A tree of claims gives every claim, and every link between them, a place you can point to — so Maren's suggestion has somewhere to land. · What a tree is, said plainly; necessity in two sentences ("to have this, we need that"); the real 2R goal tree growing from its root (animation); Maren's card landing as a missing condition under "small dense groups". Link to A5 for the method.
4. Why nobody does the filing — Structure takes effort, so it looks like a choice between easy and cumulative — but that rests on one assumption: that the person contributing has to do the filing. · The stairs; the dilemma in words, no cloud diagram; the assumption; "no longer true in the age of AI." Link to A7.
5. Talk first, file later — AI reads the conversation afterwards and proposes where each point belongs; people decide what goes in. · The worked example from Demo C: a real AI reading of rehearsal contributions against the real tree — e.g. Maya's "the goal is too broad to choose by" proposed as a challenge to the top goal, with its confidence and an alternative reading; one contribution the tree can't place yet. Stated plainly as rehearsal material and a prototype. One sentence on the tree pointing the forum back at earlier threads. Close: the status, the invitation ("find the claim you disagree with, or the one nobody is working on").

Series A — Why fixes don't stick (8 chapters)

Running example: Ashfield clinic (composite, arithmetic checkable). The open question that carries A: Dr Adeyemi loses two hours a day to admin — why? Asked in A2, answered by the current reality tree in A6, resolved with Maria in A8.

Part one · Find the constraint

1. Sixty people waiting — How much any system achieves is set by one constraint, and it is almost never where people are busiest. · Maria's desk (busiest, irrelevant) vs the doctor (twenty a day for four years). The clinic's goal is patients seen, well — throughput said plainly. Figure: five stations, patients flowing and piling before the doctor (animation). One line of lineage: Goldratt, the Theory of Constraints.
2. Squeeze before you spend — Before paying to expand the constraint, get everything you can out of it — and expect it to move once you do. · The one hire at the front desk makes a longer queue; the doctor's day has two hours of admin; give them back and the same day sees twenty-six, for nothing — but why is the admin there? (left open). Ends with the five focusing steps as one small picture: find it, squeeze it, make everything serve it, expand it, go again.
3. The rule nobody can see — The stubborn constraint is usually a rule, and because everyone experiences it as normal, five sensible people will give you five different answers about what's wrong. · Invisible rules (no pile of work in front of a rule); five staff, five confident causes. The hinge into Part two: you need a way to lay the reasoning out.

Part two · Think in trees

4. An arrow is a claim you can check — Every arrow in a diagram is a claim, read one of two ways — "this is needed for that" or "this causes that" — and anyone may fairly ask of it: do I understand it, is it true, is it the real cause, is it enough on its own? · A tiny tree first (two boxes and an arrow from the clinic), then the two readings, then four questions; fire needs fuel, heat and oxygen. The full eight Categories of Legitimate Reservation in a footnote. Lineage: Dettmer's Logical Thinking Process named here.
5. What should be happening? The goal tree — Start from the goal and ask what must be true for it, then what each of those needs: a tree read top-down with "in order to… we need…". · The clinic's goal tree grows from its root (animation). This is the tree B3 links to.
6. Why isn't it happening? The current reality tree — The many things going wrong trace back through cause and effect to a few root causes, and the deepest one is the constraint. · Undesirable effects at the top, arrows converging — on the countersignature rule. The open question from A2 is answered: that's where the admin comes from. Say the bridge line: the root cause the tree finds is the constraint Part one was looking for.
7. Why hasn't somebody fixed it? The cloud — A problem persists when fixing it seems to cost something else we need; name the assumption behind that conflict and it can often be dissolved. · The countersignature exists for safety; dropping it seems to risk safety; the assumption (only the doctor can check a script) cracks. Figure: the five-box cloud, simplified.
8. Walk it forward — Before acting, check that the fix produces what you want without new harm, then work out what's in the way and in what order. · Future reality tree in one picture (ready notes need forty minutes of somebody's morning — Maria's); prerequisites in a sentence. Maria sees her referral backlog and the countersignature are the same jam from two ends of the corridor. Close: the five questions as one picture; "keeping trees like this alive across a whole group is the expensive part — that's the other series" → B.

Kept, reshaped, dropped

Scale

13 chapters × ~350–500 words ≈ 5,500 words, against ~10,500 in the five current pieces. One idea per chapter, one to three figures, animation only where something changes over time (about five across both series: thread greying, flow through the clinic, goal tree growing ×2, a sentence travelling to its address).

Where v2 departs from the critique

• A's title "Why fixes don't stick" rather than "Find the constraint": Part two isn't about finding the constraint, and "why fixes don't stick" is already the site's tagline for this material.
• "Pull requests for discourse" is not used as framing in B5; it's the talk's line for a technical room. Rufus wants it back as a short aside (decision 4).

Decisions (resolved by Rufus, 2026-09-19)

1. Shape: approved. Two series, B (5) first, A (8) second. Maren is an invention, so no one arriving from the forum will recognise her: B says so plainly when she is introduced (composite, invented), and never implies she is a real member.
2. A's open question and running example: approved as drafted. The clinic, with the countersignature rule as the thread. Not independently checked by Rufus; a later review can revisit it.
3. B5 uses Demo C's rehearsal contributions: approved, labelled as fictional contributors, a real AI reading, and a prototype. Real contributions from the talk exist and can replace them later; that does not block shipping.
4. Voice calls. "Gardeners close" stays out of B. "Pull requests for discourse" comes back as a short aside in B5 (a line for technical readers, skippable), not as the framing.
5. Titles: keep "Where does the thinking go?" and "Why fixes don't stick". Rufus thought about the talk's title and subtitle but is not attached; revisit if a better one turns up.

Standing instruction: keep shipping; review again later rather than holding chapter work.