Evaluation — the finished 2R piece

Written 2026-09-19 after finishing second-renaissance-plain (beads reasoncommons-bnl.1–.3). Two reads: mine, and an independent cold read by a fresh agent told to read as a newcomer to both the movement and LTP. Purpose: decide what carries into the series form (costar-teardown.md) and arc.

What the piece is now

~1,500 words of prose, 12 figures, 11 sections. That is five Co-Star chapters' worth in one page. The voice pass worked: the cold reader found it "mostly plain and human" and quoted "She has a dissertation to write" as exactly right.

What landed

The problem, via the story. Maren's post arriving and going grey, and the research circle having the same conversation three times, landed early and clearly. A story first, the concept second, is the right order.
"A claim buried in a paragraph on page forty has no address." The sentence the reader would repeat to a friend. The whole case for trees hangs on it.
Contributors just talk; filing happens afterwards. Understood, but only half believed (below).

What didn't

The opening is manifesto, not story. "Paradigm in decline", "civilizational renewal" come before Maren. A newcomer should meet her first.
"A tree of claims" loses the reader. It promises one tree, then introduces five diagrams, then uses one. The five-sheets figure and the lineage paragraph are noise in this piece ("nothing here depends on it" — then why is it here?).
Logic bug at the payoff. Maren's comment is about place; it lands on "small dense groups make it livable", and the figure crosses out that condition's link to the goal — which says she thinks small groups aren't necessary. She never said that. Must fix (new bead).
The catch and the dilemma make one point twice. The conflict diagram adds ceremony, not understanding.
The AI turn is asserted, not shown. The load-bearing claim gets one bold line, and the small print admits matching isn't good enough to run unattended. "Defend it" is on the stairs and AI doesn't remove it. This needs a worked example (a real thread, three proposed placements — Demo B from the talk) rather than a sentence.
Overclaim: "she has changed what the movement believes" — a small group merged one sentence.
Undefined terms: "reason commons", "cumulate".

Figures

Carry the argument: thread-live/thread-lost, no-address, destination (once fixed), separate-acts.
Decoration: five-sheets; both-directions mostly repeats the map.
Confusing: compounds (labels far from their lines), five-map (40% empty paper), the-conflict (arrows read right to left; the cracks look like errors), the-stack (the crowd grid below the ground reads as a tally).
Want to move: thread-live→lost is one animation split into two stills (post arrives, replies gather, it greys into the run); three-times (rise and collapse, on a loop); compounds (staircase climbs while peaks reset); separate-acts (a sentence travels conversation → reader → tree → back to Maren); destination (card drops onto its condition); a tree growing (the talk's own TODO).
Stills captured from the scrolling pages carry that form's layout (wide empty panels, captions far from marks). New figures should be drawn for the page, as standalone SVGs, not captured from a scroller. The redrawn stack (the-stack-v2.svg) shows it is cheap.

Voice: one tension for Rufus

The cold reader flagged as insider or written-to-impress: "pull requests for discourse 🙂", "close to rumour", the gardeners close, "cultivate a commons for reason", naming Goldratt/Minto/LTP. Several of these are Rufus's own choices from the talk. Resolution I'd suggest, not decide: the talk's audience was insiders; a cold web reader is not. Keep the lineage, but in the LTP-basics chapters where it is the subject, not in the 2R story. Keep the gardeners image but earn it by stating the objection it answers. Replace "pull requests" with its plain meaning unless the reader is technical.

Recommendations

For the form (reasoncommons-bnl.5):

One idea per chapter, ~300–500 words, the teaser sentence written first. The cold reader's split produced nine chapters from this piece; that is about right for the grain, too many for one series.
Figures drawn as standalone SVGs for the page, placed next to the sentence they explain, labels on the marks. Animate only what changes over time; three or four animations per series, not per chapter.
A small "where this came from" footnote page per series instead of in-piece sources sections.
Fix the theme's phone-width clipping (GH #20) before shipping a series; every chapter will be read on a phone.

For the arc (reasoncommons-bnl.6):

There are two stories here and they want two short series, cross-linked. The 2R story (the problem and the Reason Commons idea: lost contributions, no address, a tree, talk first / file later, the map talks back), about 5–7 chapters, is motivation. The LTP basics (goal, constraint, the trees, the rules for arguing about arrows) is the method, following the book, and is where the lineage belongs.
The 2R series should end on a worked AI example, not an assertion. If the demo isn't ready, say so and end on the tree.
Maren is a good running character. The clinic (from The Wrong Queue) is the candidate running example for the basics.

Immediate fixes to the current piece (new bead, small): fix the destination logic, cut the five-sheets and lineage paragraphs, start with Maren, soften the overclaim, define "reason commons". Worth doing because the piece stays live until the series replaces it.