# 2026-03-29 — Session Log

## Spoken New Site — #new-site / #home-page channels

### What happened
Assaf opened new work stream: build a new Spoken website. Channel: `#new-site` (guild 1479964599099854940).

**Phase 1 — Knowledge consolidation**
- Created `work/active-projects/spoken/SPOKEN-BRAND-KNOWLEDGE.md` — master brand doc consolidating all Spoken knowledge: method (Present→Rehearse→Harmonize→Restate→What If→Create), app design system, Spoken Institute, campaign, photography direction, production assets.

**Phase 2 — Site direction agreed**
- Three-section structure: Vision → Current Success → Physics
- Current site `spokenmethod.com` has real proof: Mobileye, Check Point, Similarweb, Palo Alto, Intuit logos + testimonials (Abraham Willis, Ortal Tolone, Chunek Mayren). Stat: $490 vs $16,075 per developer. 7 years, Israeli market.
- New site leads with WHY/HOW (method, philosophy) then validates with proof. Current site does it backwards.

**Phase 3 — Hero prototyping (clou.ch reference)**
- Assaf referenced `https://www.clou.ch/` as hero inspiration
- Reverse-engineered clou.ch stack: GSAP 3.12.5 + SplitType + Lenis + Webflow IX2
- **Real mechanic** (took 3 iterations to get right):
  - Canvas layer = 15 project images, `position: absolute`, scattered via `nth-child` CSS
  - Entire `.hero-canvas` moves as ONE unit via `transform: translate3d + rotateX/Y` on mousemove (Webflow IX2 parallax)
  - "CLOU" H1 stays fixed, centered, z-index above canvas
  - Cards entrance: `scale(0.5) → full` with `back.out` ease, random stagger
  - H1 entrance: SplitType chars `yPercent: 110, rotation: -25, autoAlpha: 0 → visible`

**Phase 4 — Typography effect (Collins SF Symphony reference)**
- Assaf referenced `https://wearecollins.com/case-studies/san-francisco-symphony/`
- Effect: per-letter font-size AND font-weight randomization. Not just scale — actual size + weight mixing.
- Collins quote: "responsive and variable font technology — giving each typographic character the ability to immediately change form"
- Implementation: SplitType → each `.char` gets random fontSize from `[0.35em → 3.2em]` and fontWeight from `[200, 400, 800]` (biased to extremes). On mousemove: all chars re-randomize.

**Phase 5 — Font + cards**
- Assaf sent PP Editorial New font zip (Pangram Pangram — personal use)
- Fonts deployed to `public/spoken-new/` (6 weights: Ultralight, Regular, Italic, Ultrabold, UltraboldItalic, UltralightItalic)
- Cards reduced from 15 → 6, nicely spaced
- Hero bg changed to `#4C325A` (dark plum/purple) — Assaf's request

### Current state of prototype
- **URL:** `https://staging.curiousendeavor.com/spoken-new/`
- **File:** `work/active-projects/spoken/new-site/index.html` + fonts in `public/spoken-new/`
- Hero: `#4C325A` bg, "SPOKEN" in PP Editorial Ultrabold warm off-white, 6 cards (2 cream, 2 lime #C8FF00, 2 frosted glass), parallax mousemove
- Typography section: dark `#0C0C0C`, variable weight+size per char on mousemove
- **Still pending:** Assaf shared a Google Drive video showing the clou.ch effect — needs access. Asked Assaf to share with kitt@curiousendeavor.com. Drive file ID: `1vt2v53Nl2Jlsc1xPl3vflGvNxw24DmjA`

### Decisions made
- PP Editorial New = font choice for Spoken new site
- Hero background: `#4C325A`
- 4-6 cards in hero (not 15)
- Variable type: mix font-size + font-weight (not just scale)
- No variable font needed — PP Editorial's weight range (200→800) provides enough contrast for the variable effect

### Files created today
- `work/active-projects/spoken/SPOKEN-BRAND-KNOWLEDGE.md`
- `work/active-projects/spoken/new-site/index.html`
- `public/spoken-new/index.html` + 6 PP Editorial .otf files
- `work/active-projects/spoken/spoken-current-site.png` (screenshot of spokenmethod.com)
- `work/active-projects/spoken/collins-sf-ref.png`
- `work/active-projects/spoken/clou-reference.png`

---

## ORT R&D Teacher Campaign — Image Generation Session (14:00–15:23 UTC)

### Context
Working in #r-n-d-teacher channel (guild 1479964599099854940, channel 1487736180765364255).
Campaign bible: `work/pitches/ort-rnd-teacher/campaign-bible.json`
Character reference (locked): `2026-03-29-ort-character-ref-v2.png`

### Character (locked)
- Man, late 30s, fair skin, light reddish-auburn ginger beard (short/neat), wavy auburn-reddish brown hair
- Cream linen button-down (sleeves rolled), slim khaki chino trousers, brown leather Chelsea boots
- Trousers came out olive-gray in ref (not warm sand) — model's natural output, not overriding

### Scenes completed today
| Scene | Status | File |
|-------|--------|------|
| Scene 3 — Teaching (Monday 10:00) | ✅ Locked on v6 | `2026-03-29-ort-scene03-teaching-v6.png` |
| Scene 4 — Mentoring (Tuesday 13:00) | 🔄 In progress (v3 rejected angle) | `2026-03-29-ort-scene04-mentoring-v3.png` |
| Scene 5 — Research (Wednesday 16:00) | ✅ Approved | `2026-03-29-ort-scene05-research-v1.png` |
| Scene 6 — Coaching (Thursday 10:30) | 🔄 Description approved, not yet generated | — |

### Key learnings from this session
1. **Always describe before generating** — Assaf confirmed this workflow explicitly. Describe → get approval → generate.
2. **Consult storyboard images before prompting** — use `image` tool to analyze storyboard angle/composition before writing prompts. Do not guess.
3. **Storyboard analysis trick:** Ask the vision model specifically about camera angle, depth planes, character positions separately — gives more precise prompt language.
4. **Scene 3 angle:** Straight-on wide shot, eye level — NOT diagonal. Whiteboard left, steps right, teacher center.
5. **Scene 4 angle:** Ground level, one-point perspective, strong convergence lines (floor tiles + ceiling both converging to vanishing point at characters).
6. **Students/people:** Must always be "content and engaged, not intense" — model defaults to confrontational. Explicitly add: "warm, genuinely curious, slight smiles, NOT angry, NOT confrontational."
7. **Teacher pose:** Model defaults to symmetric "arms open" stance. Break it explicitly: "weight shifted onto right foot, torso rotated, one arm making a gesture — asymmetric, NOT symmetric."
8. **Scene 5 — Research:** Round table = equality. Teacher is the calmest person in the room. Lab coat on chair back. Molecular model on table.
9. **Scene 6 brief:** 3D printed prototype objects (matte plastic, geometric) on the work table — NOT metallic gears/industrial machinery.

### Scene 6 description (approved, ready to generate)
- Makerspace/fab lab, double-height, steel columns, factory windows, workbenches
- Camera: medium-wide, slightly elevated, looking down at central work table
- Table: wooden, blueprint spread flat, 3D printed plastic prototypes on top (drone parts / architectural models / robotic joints — matte, geometric)
- Teacher: LEFT, leaning over table, both hands on surface, one finger pointing at blueprint
- Young woman: center-left, leaning in, dark ponytail, grey-teal top
- Young man: center-right, arms crossed, observing, orange t-shirt
- Young person: right, holding 3D printed component, examining it
- Background: CAD monitor left, 3D printer right, technical diagrams on walls