Seasonal Intelligence: restructure into A Day's Walk + Seasonal Intelligence

Theory revised and narrowed to what the evidence supports; renamed The
Relevance Theory. Project split into two books and six essays, with the
three-tier relevance model promoted to the flagship.

- EVIDENCE.md   claims sorted into four tiers, governs what may be claimed
- CRAFT.md      lyric/landscape register, governs how it may be said
- PORTFOLIO.md  the two-book plus six-essay split, and what was cut
- BOOK_ONE_CHAPTERS.md  walk spine on the Stezka Ceskou Kanadou,
                        beats for all 14 units, argument ledger
- FIELD_BRIEF.md        what to capture on the trail
- drafts/       prologue, ch 3, ch 5 (register tests, specifics are placeholders)
- archive/      original theory and outline, unchanged

Cut: the foreign-policy argument, the weight-loss protocol, the diagnostic
scoring, the I/me labels, the seasonal network mapping (demoted to metaphor
in Book Two).

CRAFT.md

@@ -0,0 +1,99 @@
+# Craft rules — *A Day's Walk*
+
+*`EVIDENCE.md` governs what you may claim. This file governs how you may say it. Between them, they should make it very hard to write a bad page.*
+
+**Register:** lyric / landscape. Macfarlane, Solnit, Lopez, Jamie. Long sentences that earn their length. Close physical attention. Argument arriving sideways, through image, rather than announced.
+
+**The one-line test:** if a paragraph could appear in a magazine article about attention, cut it or rewrite it. This book should be unmistakably one person's, in one landscape, over eight days.
+
+---
+
+## The five rules
+
+### 1. Scene before claim. Always.
+No chapter opens on an abstraction. Every chapter opens with your body somewhere specific — light, weather, ground underfoot, what you can see, what hurts. The idea arrives after the reader is standing somewhere.
+
+The failure mode, and it's the one to watch for constantly: *"Modern information environments overwhelm our capacity to filter."* Nobody is anywhere. Nothing happens. That sentence should never survive a draft.
+
+### 2. Evidence lives in the middle third.
+Open in scene. Move to thinking. Bring in a study, or a historian, or Kahneman, only once the reader is already inside the question. Then leave — return to the ground, the light, the walking. Never end a chapter on a citation.
+
+Rough shape: **40% scene · 40% thinking · 20% evidence**, and the evidence is a guest, not the host.
+
+### 3. No apparatus in the body.
+No bullet lists. No sub-headings inside chapters. No numbered frameworks. No tables. Everything the outline expresses as a list must become prose or must go.
+
+The three tiers are the obvious danger — they are a list, and the book's core idea. They must never appear as a list. They appear as three days of walking, and the reader assembles the model themselves. That is the whole trick of the book and it is worth the difficulty.
+
+Citations to endnotes. Never in-text parenthetical.
+
+### 4. Hedge in the first person, never in the passive.
+Tier C is most of Part II. Hedging is mandatory. But *"it may be hypothesised that relevance is sorted by proximity"* is academic throat-clearing and kills the register instantly.
+
+Write instead: *"I think this is what the sorting is for. I can't prove it. Nobody has tried."*
+
+First-person uncertainty is a **novelistic asset**. A narrator who admits the edge of what he knows is more trustworthy and more interesting than one who doesn't. Every hedge is a characterisation opportunity. Use them.
+
+### 5. The reader must always want the next page.
+Nonfiction usually fails this because the reader can see the argument's destination from chapter two. The walk fixes it structurally — there is a route, weather, a body getting tired, a phone that will eventually come back on. Something is always unresolved.
+
+End every chapter on movement or on a question, never on a summary. If a chapter's last line could be replaced with "in the next chapter we will see," rewrite it.
+
+---
+
+## The walk
+
+**Eight days.** Continuous, chronological, no flash-forwards. The reader is on the walk the whole time; the thinking happens while walking, as thinking actually does.
+
+Use **your** walk — Czech or central European terrain you have actually crossed. The specificity is not decoration; it is the entire authority of the book. Real place names, real weather, real species, real people met. A reader can tell invented landscape from remembered landscape within a paragraph.
+
+**Two flashback permissions only,** used sparingly: the desk (what you left) and one earlier walk. Nothing else. Every additional timeline costs the reader their footing.
+
+### Recurring objects — plant early, pay off late
+
+- **The phone.** Off, or out of signal, for most of the walk. Its return in chapter 6 is the book's biggest scene. Mention it early enough that the reader forgets it, then feels its absence, then feels its return.
+- **The map.** Paper. What is on it, what isn't. What "a day's walk" looks like as a measurable distance — the book's title made physical.
+- **The ridge.** Seen from below on day two, reached on day four, remembered afterward. The horizon as an object rather than a metaphor.
+- **The people met.** Three or four, briefly. They carry tier two — everything you learn about places you will not go, from people who have been there. Do not turn them into interviews.
+- **Hunger.** Its strange absence on the walk. Seeded around day three, returned to in chapter 12 against the memory of the desk.
+- **Talking aloud.** Plant it in the prologue, barely. Pay it off in the final paragraph of chapter 12.
+
+---
+
+## The Part IV problem, and its solution
+
+Chapters 10 and 11 are the practical ones, and in a lyric book they are exactly where the whole thing dies — the moment the prose turns to *"try asking yourself: is this within a day's walk?"*, the book becomes a different, worse book, and the reader feels the switch like a change in cabin pressure.
+
+**The solution: never instruct. Narrate the attempt.**
+
+Part IV is not "here is what to do." It is *the months after the walk, trying to keep something*. What you tried. What lasted a fortnight. Where you failed and why. The reader extracts the practice from watching someone practise, which is how novels have always transmitted instruction.
+
+Second-person imperative is banned from the body of the book. If a reader wants a list, the appendix has one — and the appendix is allowed to be plain, because it is openly a different document.
+
+---
+
+## Banned
+
+- The self-help voice. Anything beginning *"We live in a world where..."*
+- The TED cadence — short punchy sentence, dramatic line break, rhetorical question.
+- **"Studies show."** Name the person, the place, the year: *a Stanford group put people in a scanner after a ninety-minute walk.*
+- Cod neuroscience. *"Your brain on nature."* *"Rewiring your neural pathways."*
+- Complaining about smartphones as such. The book's whole point is that the problem is older and stranger than the device, and any paragraph that reads as generic technology grumbling undermines the argument it appears to support.
+- Nostalgia for pre-industrial life. Their winters were hungry. You are making an argument about structure, not about a lost paradise, and the book is more persuasive every time it says so.
+- Chapter titles that explain themselves. *Within a Day's Walk*, not *Chapter 3: Local Information and Actionability*.
+
+---
+
+## Sentence-level
+
+- **Specific nouns.** Not *bird* — jay, or whatever it actually was. Not *tree* — beech. Wrong-specific is worse than general, so check.
+- **Verbs over adjectives.** Landscape writing dies of adjectives.
+- **Vary sentence length hard.** Long, long, long, short. The short one lands.
+- **Resist abstraction nouns**: *relevance, salience, cognition, information*. Each one is a small tax. Spend them deliberately. In any given paragraph, one is plenty.
+- **You may say "I don't know."** Often. It is one of the register's best moves and almost nobody in this genre uses it.
+
+---
+
+## Length
+
+70–80k words across 14 units, so roughly **5,000–6,000 words per chapter**. That is long enough for a real scene, a full argument, and a return — and short enough that a chapter is one evening's reading. The prologue and conclusion run shorter, 2,000–3,000.