# Caldura Design Log A running journal of design cycles — explorations and studies, and the summarized rationale for where each one landed. This is the **light tier**, deliberately less formal than an Architecture Decision Record (`decisions/`). An ADR carries a full burden of proof for a *committed* choice with system consequences. A Design Log entry captures an *exploration*: what was tried, what the current read is, and why — without pretending the question is closed before it is. ## How the tiers relate - **Design study** — the exploration package under `studies//`: comp artifacts, a viewer (`index.html`), and a per-study `JOURNAL.md` blow-by-blow. Messy on purpose; it preserves the search path including rejected directions. - **Design Log entry** (this file) — the one-screen outcome summary for a study: what was explored, the current read, a short rationale, a status, and links. The per-study `JOURNAL.md` is the detail; the entry here is the digest. - **ADR** (`decisions/`) — the committed decision with full rationale. A cycle **graduates** to an ADR once its choice locks and touches the system (tokens, generators, shipping assets). The Log entry then flips to `landed` and points at the ADR. **Status vocabulary:** `open` (exploring) · `landed` (decided — see ADR) · `superseded` (replaced by a later cycle) · `parked` (paused). ## Adding a cycle 1. Put the exploration in `studies//` (artifacts + `index.html` + `JOURNAL.md`). 2. Add a row to the index and an entry below, starting at status `open`. 3. Keep the entry to a digest — push detail into the study's `JOURNAL.md`. 4. When it locks, write the ADR, set status `landed`, and link it. --- ## Index | Cycle | Dates | Status | Study | One-line outcome | |---|---|---|---|---| | [Parent circular logo](#parent-circular-logo) | 2026-05-21 → 07-04 | `landed` | [studies/parent-circular-logo-study](studies/parent-circular-logo-study/) | **Ring mark F live on caldura.com** (F as-is; per-mode `parentMark` light+dark; favicon stays crucible-C). All 4 decisions locked + build layers L1–L8 done; sites consumed in `df310f9` ([rollout](studies/parent-circular-logo-study/rollout/index.html) · [runtime proof](studies/parent-circular-logo-study/rollout/runtime-proof.html)) | | [Establishing v4](#establishing-v4) | → 2026-05-11 | `landed` | [systems/v4/comp-sheets](systems/v4/comp-sheets/) | Adopted v4 variants selected from a top-level field (warm steel / warm distinctive / blue-institutional / AI default) vs the v3.1 baseline; C-dot parent rejected; baseline locked | --- ## Entries ### Parent circular logo - **Status:** `landed` — all four decisions locked, build layers L1–L8 done; the ring mark is live on caldura.com (sites `df310f9`, 2026-07-04). - **Dates:** 2026-05-21 (initial field) → 2026-06-04 (mark + token model locked) → 2026-07-04 (published + consumed; landed). - **Study:** [`studies/parent-circular-logo-study/`](studies/parent-circular-logo-study/) — open `index.html` for the viewer; `JOURNAL.md` for the full round log. **Question.** Keep the Caldura parent mark circular while removing its failure modes — breast/anatomy, eye/pupil (especially awkward beside AI), bullseye/target, and generic faceted-infrastructure reads — and fix a second problem the context proof exposed: the current mark goes weak below lockup scale, collapsing into a vague warm dot at nav/favicon sizes. **Explored.** A broad opening field of circular options, then ten cycles narrowing from soft/atmospheric blooms → hard-vector contour rings → true-C halo geometry → complete 3- and 4-ring systems. That last reframe shifted the evaluation from "defend against anatomy reads" to "judge as a parent mark," surfacing a four-way shortlist (A/G/D/F), a context stress-test across real sizes, and an F-focused refinement branch. The F mark was then carried into a business-card application. **Current read.** - **F · Outer Heavy, Inner Air** — strongest *technical* candidate. Survives the widest range of sizes (32/24/16/12px); outer ring carries silhouette, inner air gives depth without a plain dot-in-circle. Risk: drifts toward instrument/gauge; needs warmth tuning. - **G · Banded Source** — strongest *brand-feel* candidate. Warmest, closest to current Caldura. Risk: flattens to a warm coin at the smallest sizes. - **A · Outer Authority** — safest benchmark. Serious, parent-level, but less distinctive. - **D · Stepped Four-Ring** — likely drop. Works large, fails small as target/system mechanics. Kept for reference. - **Current** — emotionally useful baseline, mechanically weak below lockup scale. **Rationale (summarized).** The decision criterion that did the most work was small-size durability, not large-stage beauty — the original problem (anatomy + weak favicon) lives at 16/12px and OG scale, so the context proof at real sizes, not the 168px stage, is the surface that ranks candidates. F wins there; G is the emotional hedge; A is the fallback; D is eliminated by the same small-size test that promotes F. **Rollout (2026-06-03).** The cycle entered its rollout phase — a decision walkthrough that locks the skipped locks before propagating the mark up the stack ([rollout viewer](studies/parent-circular-logo-study/rollout/index.html)). **Decision 1 — Lock the mark: resolved → F as-is.** F is the parent mark, taken verbatim from the shipped card geometry (6 circles). A tune to drop F's two interior hairlines (a residual bullseye/gauge read) was rendered side-by-side and **declined**: the concentric rings are wanted for depth / crucible-aperture feel, and small-size durability is identical either way. The only argument for the tune was an optical taste call, not a failure mode. Geometry is now frozen. **Decision 2 — Tokenize: resolved → C1 (2026-06-04).** The mark gets an explicit per-**mode** `parentMark` palette — six role tokens (`ring · discOuter · discInner · hairlineMid · hairlinePale · core`), distinct from the text `accent` tokens — with values authored per mode, *not* derived by formula (the hand-tuned light hexes don't cleanly derive from `accent`; this matches the AI variant's per-mode token pattern in `adopted-variants.mjs`). Light mode = the six locked hexes; **dark mode** and the **favicon scale-variant** are authored by Decisions 4 and 3, which ride on it. Print (CMYK / Pantone) is a color-space target of light mode owned by the card spec — not a token. *Vocabulary (standard across the rollout):* **mode** = ground tone (light/dark) · **scale variant** = size floor (full/favicon) · **application** = where it ships (web/deck/card/print) · **color space** = encoding (sRGB vs CMYK/Pantone). **Decisions 3 & 4 — resolved (2026-06-04).** *Favicon → keep the crucible-C* (A): at 16/12px the C reads as a distinct letterform while the ring collapses to a generic target, and the C is the decoupled *family* favicon — no reason to overturn it. The ring stays logo / lockup / OG only, so the palette needs no favicon scale-variant. *Dark mode → author both modes* (A): the proposed dark values read clean on `#1A1210` (bright gold ring `#C79A40`, near-ground warm discs, bright core `#E3BD5C`), and a runtime lockup already points at a dark parent mark. **All four rollout decisions are now locked — geometry + tokens frozen; build layers L1–L8 are mechanical.** **Open / next.** - **Build layers — L1–L6 done.** Foundation ring (L1), generator + per-mode tokens (L2), regenerated SVGs + validator + cross-platform raster pass (L3/L4), comp-sheet + viewer dark coverage (L4), card loop closed (L5), docs (L6: ADR-0007, CHANGELOG, architecture, this entry). **Clean-before-land done (2026-06-07):** the `variant-parent-light` → `variant-parent` rename (folder + identifier; mode-encoding filenames kept) and the `bloom-mark-height-comparison` diagnostic relabeled AI-only. **L7 runtime publish done (2026-07-03,** heat `ba00307`**):** both publish scripts ran — runtime OG slots (`sites/public/brand/og/og-parent.*`) now carry the ring mark (favicon unchanged per Decision 3) and the full v4 mirror synced. Verified in the [runtime proof](studies/parent-circular-logo-study/rollout/runtime-proof.html) — nav / tab / OG in situ across all three sites. **L8 sites consumption done (2026-07-04,** sites `df310f9`**):** caldura.com consumes the ring mark (nav logo, layout, globals) per the H1 handoff in `sites/docs/heat-feedback.md` (item cleared). **All eight layers done — landed.** - **Business card (follow-up).** Direction B locked at MOO Standard; awaiting CMYK proof + PDF/X export path before the card application closes out. **Recorded in.** [ADR-0007](decisions/0007-parent-mark-f-and-per-mode-tokens.md) — the four locked decisions and rationale (status `accepted`; implementation L1–L8 complete 2026-07-04). The [runtime proof](studies/parent-circular-logo-study/rollout/runtime-proof.html) shows the published assets in situ across all three sites. A short ADR or spec for the business-card standard follows once the card lands. --- ### Establishing v4 - **Status:** `landed` — the adopted v4 baseline the current system documents. - **Dates:** → 2026-05-11 (v4 strategic brief / architecture). - **Artifacts:** [`systems/v4/comp-sheets/`](systems/v4/comp-sheets/) — top-level variant comparisons, the rejected C-dot direction, and the locked-lockups reference. Surfaced as a cycle in [`studies/index.html`](studies/index.html). **Question.** Which brand-system direction does Caldura v4 adopt, across the parent, Capital, and AI entities? **Explored.** A top-level variant field — warm steel, warm distinctive, blue-institutional, and AI default — measured against the v3.1 baseline, plus a parent C-dot direction explored on its own. **Outcome.** Adopted **parent-light**, **blue-institutional** Capital, and **amber** AI. The **C-dot parent direction was rejected**, and the accepted lockups were locked as the reference baseline the current system builds on. **Note.** This is a retrospective digest. The supporting comparison sheets and rejected directions were factored out of the brand-artifact viewer into this cycle so the viewer documents the *current* system and the cycle records *how we got there*.