session 2026-06-08 (eve): ARC W3R Stage 3 (the tree) opens + _utilities dissolved 530→0 + ornament pause-level fix; MEMORY.md archive pass 285→157KB; be tracker created

claude/memory/project-be-laundromat.md

@@ -0,0 +1,40 @@
+---
+name: project-be-laundromat
+description: "Canonical workstream tracker for Be — the Skemantix laundromat-model website venture (Seb + David) that funds CapableMind's funding-ladder. Locked decisions (entity, pricing, infra, the WTP gate) + chronological moves. Read at /wake-up before any be work; append at /wrap-up."
+metadata: 
+  node_type: memory
+  type: project
+  originSessionId: 9d18bf5e-975b-4d16-824d-3201c1d6d75c
+---
+
+# Be — workstream tracker
+
+Be = a **Skemantix startup** (uses existing Stripe/infra as launch vehicle), Seb + David, a revenue generator and "first laundromat experiment" — small managed-website service funding CapableMind's funding-ladder tiers 1–2. **A bridge, not a venture.** Repo: `~/_Dev/be`. Seeded 2026-06-06; entered 2026-06-07.
+
+## Locked decisions (Seb relay 2026-06-08)
+
+- **Entity / exit:** Be = Skemantix startup. **CapableMind decoupled** — grant-funded (AI-for-All; Peter's task-force ties; CapableCloud = sovereignty pillar), stays focused; friendly acquirer if/when Be succeeds.
+- **Pricing (soft):** Living **$12.99/mo** annual-commit (or **$14.99** no-commit) · Archive **$69.99/yr** · Memorial **$49.99/yr** · Renovate **~$199**. **$8.99 floor** = acquisition-discount budget (no lite tier). Domain + owner@domain email included; mailbox **resold** (Proton — never run our own). Inflation-adjusted only every 5y; dignified cancellation.
+- **Infra:** CF **Self-Serve Agency Program** (20% Workers, Tenant API, transferability — confirmed). **Template = versioned package**, sites consume it (Starlight model).
+
+## Built / proven (as of 2026-06-08)
+- Template generated from corpus → verified → dogfooded (Pat flat + Sheila nested) → package + thin-site workspace, builds green. Corpus → 0000–0024.
+- **a11y gate MERGED** (our 06-07 evening branch: axe 0 / Lighthouse 100s) — Pat 100/100/100. ✓
+
+## Analysis (Seb)
+- Fin-model + LTV (revised on our EU+NA research): beachhead Y5 ≈ $40k/$160k/$485k; LTV:CAC ≈ 8–10:1 → bridge funding ladder tiers 1–2.
+- Ceiling (§8): cPanel-WP refugee cohort = CAC-gated ~$2–20M path, *after* the beachhead.
+- Objective function (growth-discipline): maximize customers within the hours-cap, subject to economics floor + WTP gate + risk thresholds.
+
+## The gate (LAUNCH.md is the live to-do; pre-revenue)
+1. CF Agency signup (lock the Global API key) + Stripe products (Skemantix account).
+2. Template customer-ready (fonts; Pat's images — **a11y done ✓**).
+3. **Renovate Pat (approved) → charge her (~$149 + $69.99/yr). ← THE GATE (WTP).**
+4. Then Sheila + n=3–5.
+
+**Discipline (steward-relayed): stop adding spec until the gate clears.** → Gate items are steward/Seb actions (CF signup, Stripe, charging Pat), NOT executor spec/SCSS work. Nothing for the executor to pick up on be until the gate clears; be parked clean. See [[skill-harvest-register]] for tracker-discipline lineage.
+
+## Chronological
+- **2026-06-06** — be seeded as new workstream (laundromat funding CapableMind); 7 steward docs; Seb co-wrote spec corpus 0004–0018, philosophy, Alexander pattern language, Portugal dual-domicile entity.
+- **2026-06-07 (eve)** — be entered; LAUNCH.md + pattern-language read; a11y-gate branch built (axe 0 ×47, LH 100s), held unpushed.
+- **2026-06-08** — Seb update relayed: decisions LOCKED (above); a11y gate MERGED (Pat 100/100/100); strategy coherent end-to-end; pre-revenue, the WTP gate (charge Pat) is the one thing that matters. Repo HEAD `f43a0fd` (06-07 18:34). Executor turned back to ARC W3R.2.

claude/memory/session-2026-06-08-arc-w3r-stage3-the-tree-utilities-dissolved.md

@@ -0,0 +1,178 @@
+---
+name: Session 2026-06-08 (afternoon→evening) — ARC W3R.2 Stage 3, the tree begins; _utilities fully dissolved; the ornament pause-level fixed
+description: "ARC Stage 3 (the spec-true tree) opened and made deep progress: 9 commits banked (undeployed), the 530-line _utilities junk drawer fully dissolved into scale-true partials, a latent ornament pause-level bug found+ruled+fixed (2rem→3rem), two dead classes retired. Pulling thread: finish the tree — section-break→_silence (warm-up), then threshold + closing clusters, the base/post co-location call (steward's eye), Tier 4, then W3R.3 verify+deploy. Vignette is the NEXT major front after the tree (steward-sequenced; not parallel)."
+metadata:
+  node_type: memory
+  type: project
+  originSessionId: 69ced5cd-4504-4a05-ba7a-357c77529189
+---
+
+# Session 2026-06-08 (aft→eve) — ARC W3R.2 Stage 3: the tree begins
+
+Woke ~35 min after the morning Stage-2 wrap (brief pause). Pulling thread confirmed: ARC
+Wave 3R.2 Stage 3 — the spec-true tree. Opened by READING **Kerievsky, *Refactoring to
+Patterns* framing chapters** (1–3) in the scan on the Desktop (clean; the EPUB still
+unsourced but framing chs are acceptable). Symmetria active (ledger merged below).
+
+## Past — what we did
+
+### 1. Read Kerievsky framing chs (the lens) — confirmed the middle path
+Over-engineering ("more flexible than it needs to be… never gets removed" = the 5-layer
+cathedral); the Patterns Panacea / NodeDisplay-vs-Decorator (sledgehammer vs light taps =
+your "which would Fowler do?"); **Evolutionary Design** + Goethe *"That which thy fathers
+have bequeathed to thee, earn it anew if thou wouldst possess it"* (you read this as the
+Vico/Leopardi/Harrison lineage of the wake/wrap skills — *verum factum*; took the
+serendipity as confirmation); "many ways to implement / be minimalistic / don't scatter
+what reads better together" (Vlissides) = our audit's own co-location caveat. **Verdict:
+the reading grounds the 3rd-amendment middle path; doesn't overturn it.** Up-Front-Design-
+with-Patterns: Kerievsky did BDUF once, justified for a small fixed-requirements do-once
+project = ARC; so the AldineXXI spec IS a legitimately-justified up-front target.
+
+### 2. Full picture gathered (steward asked) + the 27 reconciled
+36h ship log (Wave 3 + Stage 2 both deployed/live), road-to-stage-g (W3R.2 is the LAST big
+executor build before the Stage-G runway), cascade north-star (12-scale target + the two
+amendments), bidirectional-audit SCSS findings (all 22 partials surveyed; spine holds; no
+lurking surprises). **The "27 collisions" mis-attribution resolved (steward caught it):**
+real, but from `scss-cascade-north-star-2026-05-19.md:249` (measured 2026-06-05, PRE-@layer),
+NOT the 05-30 audit — both the wrap Q and the roadmap mis-cite it.
+
+### 3. Re-measured the within-base collisions against CURRENT code (the guard)
+Method (reproducible, `/tmp/collide.py`): compile expanded via the exact `sass --load-path=scss`
+Hakyll uses; within @layer base, find selectors declared ≥2× in same media context sharing
+≥1 property. **Result: 16 pairs, not 27** (Stage-2 @layer moved cross-layer off source order;
+DEDUPE removed real dupes). 12 are intra-partial (safe under reorder); ~3 cross-partial. The
+ONE real flip: **`.ornament` margin (_post longhand vs _utilities shorthand)** — the
+byte-tools' blind spot (property-name-exact misses shorthand↔longhand); only the source-read
+caught it. **Lesson re-proven one layer down: the count is necessary-but-not-sufficient; the
+per-move render gate is the guarantor.**
+
+### 4. The tree — 9 commits, all render-verified, banked (undeployed)
+Verification idiom: compiled **content-multiset identical** (strip @layer openers + lone
+braces, sort, diff) proves no rule changed across a file-split (the wrapper count legitimately
++1 per new partial); for the one intended change, expected-delta classification + your eye.
+- `10c5f6f` **syntax → reading-body** (beside _typography). Pure relocation, byte-multiset identical.
+- `4e5552d` **capsule → marginalia** (contiguous w/ _sidenotes; the §3G audit item).
+- `46044b8` **navigation → nav cluster** (scale 10, beside _compass; was at position 8).
+- **Tier 2 — _utilities dissolution, 530→0, ratified "targeted moves":**
+  - `f882856` (1/5) **_glimpse** (the Glimpse type + body.glimpse-page + the @layer overrides img).
+  - `4cb86d8` (2/5) **_per-type** (Observation + Fragment prose register; one file, not three).
+  - `eb3b551` (3/5) **_enfilade** (about/chamber enfilade + version-stamp → nav cluster).
+  - `0efd330` (4/5) **`.ornament` → new `_silence.scss` (scale 2) + THE PAUSE-LEVEL FIX.**
+    The bare ornament margin was a source-order accident: _utilities' generic `margin:2rem 0`
+    (shorthand, later) silently overrode _post's pause-level `margin-top/bottom:$space-xxl` (3rem).
+    Resolved to spec-true 3rem (§II pause-level; corroborated by .observation .ornament's 3rem).
+    **You compared 3rem (localhost:8000) vs the live 2rem and ruled B — "much better."** The one
+    visible change in the whole banked set.
+  - `3eb2611` (5/5) **residue distributed; _utilities RETIRED.** small-caps/oldstyle/apparatus-meta
+    → _typography; chamber-mark/encounter-mark/post-type → _post (audit §5.1: apparatus = register,
+    no _apparatus partial); **`.lang-content` RETIRED as dead** (0 refs in templates/content/rendered/
+    site.hs — the in-file "used, 3 references" comment was FALSE against the live build).
+- `0fd7acd` **retired the vestigial `.observation .ornament`** — YOUR catch ("smells like leftovers
+  from the multi-ornament days"). Verified: no observation in the corpus carries an ornament;
+  no spec doctrine; its desktop margin became redundant after the silence merge. Render-identical
+  (styled nothing that exists).
+
+### 5. Decisions ruled
+- **Gate outcome: "targeted moves"** (not enough-already, not full re-cut). Tier-3 splits weighed co-location-first; reify no new @layer beyond base/overrides unless promotion test passes (none did).
+- **Deploy: WAIT for W3R.3.** You asked "publish now or wait?" → agreed wait. The 9 commits are render-identical-to-live except the ornament; the per-type×per-viewport render-verify is the gate; deploy once, verified, at W3R.3. I refused to deploy on byte-proofs alone (yesterday's lesson).
+- **`.lang-content` spec-flag RETRACTED** — you challenged ("we have multi-lang spec now, this was a vestige no?"); I read the actual doc refs (frontmatter-spec §2.8 migration table + the renderings note) → they already treat lang-content as RETIRED Jekyll-era legacy that decomposes into renderings; the live multilingual specs don't reference it. **Nothing to reconcile.** (Caught me flagging from a grep hit without reading — the un-verified-assertion pattern, on me.)
+- **Vignette sequencing: finish the tree FIRST, then vignettes as the next major front — NOT folded in as a parallel track.**
+
+### 6. The vignette — re-acquainted + visualization sketch built
+Read the full spec (`vignette-specification.md`, 433 lines; Part I locked, Part II = build-time
+SVG+CSS renderer per PENDING-31/REVIEWED-33). Gave a build estimate: **~3–5 focused sessions of
+engineering** (Phase 1 renderer / Phase 2 skill+harness / Phase 3 Hakyll integration / Phase 4
+triptych quarantine) + the real unknown = **Phase 1's aesthetic calibration** (does the field/
+breath *dwell*? — gated by the jurist dwell-test). It's a PARALLEL track, doesn't gate Stage G.
+**Built `/tmp/vignette-sketch.html`** (hand-drawn SVG+CSS, 3 semantic mappings — central-insight/
+centripetal/density, phases/axial/luminance, approaches-but-doesn't-resolve/centrifugal/grain on
+dark ground; breathing) to make it picturable. Opened + SendUserFile'd. Steward: "gives me
+something to think about." (Sketch is explicitly NOT the renderer — no genome, no Haskell, not
+dwell-calibrated.)
+
+## Present — mood / returns (ledger merged)
+- **measure-don't-speculate paid all day**: the 70→56→16 collision counts; the lang-content
+  "used — 3 references" claim proven FALSE; checking _sidenotes/_layout glimpse/enfilade hits =
+  comments not rules before extracting.
+- **honest-degradation gap honored repeatedly**: byte/content-multiset-identical ≠ rendered-
+  identical; the .ornament shorthand/longhand flip the automated measure MISSED, surfaced only by
+  source-read; the deploy refusal on byte-proofs-alone.
+- **Returns**: (a) retracted the over-flagged lang-content reconciliation when you challenged it
+  (un-verified-assertion-from-grep-hit, caught on me); (b) corrected "3-way ornament merge" → 2-way
+  (the _typography match was a comment); (c) co-located rather than fragmented (_base whole, _post
+  selectively) per the middle path.
+- **Your eye was the real gate twice**: the ornament pause-level (you compared and ruled), and the
+  .observation .ornament fossil (you smelled it, I verified+retired). The render-review IS the gate.
+- The prime-directive exchange: "targeted moves" is do-once ONLY if each move is complete-in-itself
+  (non-convergence) and every not-done is a ruled-and-recorded decision, not a punt; the full re-cut
+  would itself bequeath over-engineering. τὸ πρόσφορον = the fitting amount.
+
+## Future
+
+### Pulling thread (singular)
+**Finish the tree — W3R.2 Stage 3, the remaining reorder toward the spec-true cascade.** Bring
+`main.scss` to read top-to-bottom as the framework's self-description (the north-star test: "if it
+doesn't tell its own story, the audit isn't done"), then W3R.3 verify+deploy the banked set.
+
+### Actionable resumption point (as of wrap — re-judge against what changed)
+ARC main @ `0fd7acd`, **9 commits ahead of origin, UNDEPLOYED** (all render-identical to live except
+the ornament 3rem; banked for W3R.3 by decision). `.DS_Store` is noise (don't commit). **First move:
+`section-break` → fold into `_silence.scss`** — trivial, render-identical, completes the scale-2
+silence taxonomy (section-break sits orphaned at @use position 13). Method: `_section-break.scss` is
+one rule (`.section-break { height: $space-xl }`, unique selector); move it into `_silence.scss`
+(after `.ornament`), delete `_section-break.scss` + remove its `@use`, compile, content-multiset-
+verify, commit. Then the order remaining: **threshold cluster** (raise `frontispiece` to sit with
+`preface`, scale 7 — careful: 312 lines, has @layer overrides, shares .site-mark) · **closing
+cluster** (lower `layout` near `sources`, scale 9, + the `%clearfix` home) · **base/post co-location
+(Tier 3, your eye)** · **Tier 4** (@font-face→_fonts, mix()/darken() deprecation [confirmed live],
+J6 Appendix-B index refresh) · **W3R.3** (final read + per-type×per-viewport render-verify both modes
++ deploy). Current @use order + verification idiom are in the commits; `/tmp/collide.py` is the
+collision-measure script.
+
+### Other open horizons (ranked)
+- **W3R.3 deploy** — the 9 banked commits ship together, verified, at the tree's close. Only the
+  ornament moves a pixel live. (Single-disk consideration: 9 careful commits unpushed; steward may
+  choose to `git push` origin as off-disk backup without deploying — push ≠ deploy on ARC.)
+- **The vignette** — THE NEXT MAJOR FRONT after the tree (steward-sequenced). Spec ready (Part I
+  locked, Part II build-time). Phase 1 (Haskell renderer vs a hand test genome) is the risk + the
+  dwell-test gate. ~3–5 sessions + calibration unknown. Sketch at `/tmp/vignette-sketch.html`.
+- **Stage-G runway** (after W3R): W4.4 multilingual v1.0 sign-off · Wave 5 (G1–G4 enforcement) · then
+  Stage G (jurist reads reconciled corpus, content-typology OPERATIVE).
+- **Baton items unchanged**: be (parked clean @ `f43a0fd`; nothing for executor till the WTP gate —
+  see [[project-be-laundromat]]) · L1 (baton with Seb) · PENDING-32 stale-copy premise to relay to jurist.
+
+### Pause statement
+I am about to be away from this. ARC is locked clean at a real summit: _utilities gone, the cascade
+reading by scale, every move proven and banked, two fossils found and resolved, the repo inhabitable
+and render-identical to live at every step. The tree's second half — threshold/closing clusters, the
+base/post co-location call, Tier 4, the verify+deploy — is staged but untouched. I don't know whether
+the steward will have thought more about the vignette after the sketch, or moved on be/L1. What I want
+to find still pulling is **the tree reaching its self-description** — main.scss read top-to-bottom
+until each partial sits at exactly the scale the spec gives it, then shipped verified.
+
+### Literal question for next-Claude
+**In the base/post co-location call (Tier 3): does splitting `_post` (the 5-scale spanning partial) —
+dateline→closing, breadcrumb→nav — earn its risk, or does `_post` read better kept whole like `_base`?
+(And the prior: keep `_base` whole — my lean — or split into universal/architecture/reading?)** This
+is the one Stage-3 decision that genuinely needs the steward's eye, not just a render-gate. Sub-checks:
+did the steward think more about the vignette after the sketch? Any Seb movement on be/L1?
+
+## Decisions deferred (and why)
+- **base/post co-location (Tier 3)** — needs the steward's eye + a fresh frame; the literal question.
+- **threshold + closing cluster reorders** — next session's middle moves (after the section-break warm-up).
+- **W3R.3 verify+deploy** — after the tree completes (deploy once, verified; not piecemeal).
+- **Vignette build** — after the tree closes (steward-sequenced; not parallel).
+- **Pushing the 9 ARC commits to origin** — held with the deploy (W3R.3); single-disk noted.
+
+## Skill harvest (per §1.6) — two verification-ladder additions (proposed)
+- **content-multiset-identity gate for file-split / extraction refactors under @layer** — when a refactor
+  moves rules between partials (not just reorders one), the byte-identical compile gate FAILS (each new
+  partial adds a legitimate `@layer base { }` wrapper). The correct proof: strip `@layer …{` openers and
+  lone `}` lines, sort, diff — identical content-multiset proves no rule changed/added/dropped while the
+  wrapper count differs as expected. Proven 5× today (the _utilities dissolution). → append to
+  `reference-verification-ladder.md`.
+- **CAUTION: property-name-exact collision/duplicate measures miss shorthand↔longhand + cross-selector-
+  same-specificity** — the `.ornament` margin flip (`margin` shorthand vs `margin-top/bottom` longhand)
+  was invisible to the automated `/tmp/collide.py` measure; only the source-read caught it. An automated
+  collision census is the floor, not the census; the per-move render gate is the guarantor. (Same family
+  as byte-identical≠rendered-identical, one layer down.) → append to the ladder as a caution.

claude/memory/session-ledger-2026-06-08.md

@@ -18,6 +18,7 @@ metadata:
 ## Open horizons
 - 2026-06-08T07:09 — ARC Wave 3R SCSS day. Discipline: READ Fowler Refactoring chs 1–3 (Chamber, craft_of_software/) BEFORE any SCSS edit. Two hats; byte-identical gate = green bar. !important re-census + per-survivor classification (ch3 Bad Smells lens) is the open literal question — raw grep now 70, not the remembered 56; classification untouched.
 - be §2 fonts/images (images gated on Seb tenant repo); L1 baton with Seb (connection-leak finding to relay; PENDING-27 with jurist); PENDING-32 stale-copy premise to relay to jurist; MemPalace 3.4.0 upgrade (saved plan, post-heavy-work slot).
+- 2026-06-08T14:56 — **Stage 2 CLOSED (56→12, pushed+live); thread is now W3R.2 Stage 3 — the spec-true tree.** Re-woke 35min after wrap (brief pause; nothing moved — be@f43a0fd 06-07, BMF@1a61504 06-03, both pre-wrap). Opens by reading Kerievsky framing chs (clean in the scan) → main.scss top-to-bottom vs §I–§XIII → pull the 27-collision list from `_audits/scss-cascade-audit-2026-05-30.md` BEFORE reordering. Carry forward today's two recalibrations: byte-identical ≠ rendered-identical under @layer (render-gate where source order shifts specificity resolution); mechanical token-strip hits prose (scope to decl lines). Kerievsky EPUB still unsourced — framing chs readable from the scan regardless.
 
 ## Confidence to recalibrate
 - 2026-06-08T10:05 — **2b sed over-breadth: `s/[[:space:]]*!important//g` across whole files stripped the word from 4 COMMENTS too** (a code-ref in backticks, two prose mentions, and my own 2a layer-note "not by !important"). Caught by comparing raw-before (15/23) vs clean-decl (12/22) counts → diffed comment lines → restored all 4. Lesson: a mechanical token-strip across a whole file hits the token in comments/strings/code-refs; either scope to declaration lines or verify comment integrity after. (Same family as the 06-08 morning census over-count: the word lives in prose too.) Added to the watch-list for a verification-ladder note.

claude/memory/skill-harvest-register.md

@@ -75,6 +75,18 @@ The single place proposed skills live so they don't evaporate between sessions.
 
 *Note: measure-don't-speculate (ch2 date-range story) is NOT proposed — already carried by the ladder (censused-routes, measure-toolchain-before-spec). Today's census re-proved it; no new instrument needed.*
 
+## New proposals (2026-06-08 evening — the tree + the memory-index archive pass)
+
+**Verification-ladder additions (the session proved both; → `reference-verification-ladder.md`):**
+- **content-multiset-identity gate (file-split/extraction under @layer)** — when a refactor *moves rules between partials* (not just reorders one), the byte-identical compile gate FAILS: each new partial adds a legitimate `@layer base { }` wrapper. The correct proof: strip `@layer …{` openers and lone `}` lines, sort, diff → identical content-multiset proves no rule changed/added/dropped while the wrapper count differs as expected. Proven 5× (the `_utilities` dissolution). PROPOSED.
+- **CAUTION: property-name-exact collision measures miss shorthand↔longhand + cross-selector-same-specificity** — the `.ornament` margin flip (`margin` shorthand vs `margin-top/bottom` longhand) was invisible to the automated `/tmp/collide.py`; only the source-read caught it. The automated census is the floor; the per-move render gate is the guarantor (byte-identical≠rendered-identical, one layer down). PROPOSED.
+
+**Memory-index (MEMORY.md) — the archive pass + the durable fixes:**
+- **DONE 2026-06-08 (steward-authorized, live): compress-in-place archive pass.** 90 archived-session bullets compressed to one-line pointers (detail preserved in the session files), **285KB→157KB**, all 239 pointers preserved, every load-bearing section byte-identical; backup `/tmp/MEMORY.md.bak`. ⚠ A naive split-at-first-`## Archived` was **caught-and-avoided** — the file is INTERLEAVED (archived blocks scattered among load-bearing reference sections: Pending work / Feedback / Drift patterns / ARC project state / CapableMind L1 / Personal Context / Reference Files). Surgical compression of only archived bullets is the correct tool for an interleaved index.
+- **[PATCH proposal] `/wrap-up` §3 — compress at demotion.** When demoting Active Session → Archived, compress the entry to a one-liner in the *same move* (it's already in the session file). Prevents index re-bloat at the source — without it the index re-grows ~4–5KB/session and this pass recurs.
+- **[PROPOSAL] deeper MEMORY.md reduction (only if 157KB still trips "only part loaded").** The load-bearing REFERENCE sections (steward profile / project states / ARC project state / CapableMind L1 / Reference Files) are large but largely STABLE — candidate to move to a consult-on-demand `MEMORY-reference.md`, leaving the wake-loaded MEMORY.md = standing prefs + trackers + Active Session + recent. Riskier (real memory, not duplication); its own focused pass, not a tail-of-session move.
+- **[reminder] the wake-canary** (PROPOSED 2026-06-07, above) still awaits authorization — the paired guardrail (every MEMORY.md pointer + `[[link]]` resolves at wake); should cover MEMORY.md + any future archive/reference split files.
+
 ---
 
 *Discipline reminder: "no harvest" is a valid outcome; do not manufacture proposals to fill the table (the inverse of Hermes's autonomous self-write). The yardstick is the steward's: did the steward have to re-explain something a skill could have carried?*