Grand Diomande Research · Full HTML Reader

Claude Design Paste-Once Prompt

This is the prompt to paste into your **real browser tab** at `claude.ai/design` to get 5 visualized LUME pitch decks back in one shot.

Embodied Trajectory Systems proposal experiment writeup candidate score 18 .md

Full Public Reader

Claude Design Paste-Once Prompt

This is the prompt to paste into your real browser tab at `claude.ai/design` to get 5 visualized LUME pitch decks back in one shot.

How to use:
1. Make sure you're at https://claude.ai/design (logged in)
2. Copy everything between the `===` lines below
3. Paste into the prompt box
4. Hit send
5. Claude will generate a deck. After it returns, ask: "Now generate the next 4 variants in the same style: experience-first, cultural-first (N'Ko/Bambara), traction-first, market-first. Each as a separate deck I can flip through."

---

PASTE BELOW (everything between the `===` lines)

===
Build me a 10-slide venture pitch deck for LUME, designed (not just outlined). Use the brand spec and product brief I'm pasting below. Output an interactive HTML deck I can flip through with arrow keys.

LUME (Variant 1: HARDWARE-FIRST framing — lead with the engineering moat).

# BRAND SPEC

Colors:
- Primary yellow: #FCE94F (titles, accents)
- Deep yellow: #F4D80E (OLED-saturation alt)
- Black: #0A0A0A (text, strokes)
- Cream: #FAF6E8 (subhead, light surfaces)
- Rust: #C13B1A (cultural / N'Ko accent)
- Graphite: #1A1A1A (dark deck background)
- Fog: #2A2A2A (dark deck card surface)

Typography:
- Titles: Anton (Google Fonts), all caps, condensed bold, letter-spacing -1px
- Subhead: Caveat (Google Fonts), cursive italic, lowercase, hand-drawn feel
- Body: system-ui clean readable
- Technical labels: JetBrains Mono
- NEVER Inter, Arial, Roboto, or Times New Roman

Layout:
- 16:9 landscape, 1920×1080 per slide, scrollable vertically
- Asymmetric. Title overlaps body content by 10-15%
- Generous negative space. Don't fill every quadrant
- One artifact per slide max
- Subtle SVG-turbulence grain texture overlay

Voice rules (hard):
- No em dashes (use commas, periods, line breaks)
- Never call me "founder" — I'm Mohamed or first-person
- No AI-isms: delve, leverage, craft, seamless, thrilled, excited to share

# LUME PRODUCT BRIEF

What LUME is:
LUME is a real-time body-driven audio + visual install engine. A song plays. A body moves. The room watches both. When the rhythm catches the body and the body locks in, the wall remembers it. The room becomes a partner. It runs as self-contained hardware in venues: bars, lounges, galleries, retail spaces, brand-experience rooms.

Technical edge:
- 21 Unity components running together at 60Hz on a small mini-PC
- Multi-modal: audio FFT + body tracking + depth reproject + optical flow + GPU compute
- Original UDP wire-format suite (audio, depth, motion, point clouds), version-stable, reusable
- Computational choreography score: real-time body-music alignment metric drives visuals back into the loop. Patentable territory.
- Editor-grade tooling: F2 choreography panel, F12 calibration panel, one-click scene auto-wire. <2hr venue install.
- Hardware-software co-design: K11 mini-PC + Femto Bolt depth cam + 1920×440 IPS bar display + 31-part 3D-printed shell (ASA, Elegoo Max).
- 73-test pytest suite + Wave 1-9 architecture lineage. Production-grade.

What's shipped (2026-05-02):
- Wave 1-8 complete: audio reactor, depth reproject, optical flow, VFX bridge, fluid sim hooks, choreography scorer, motion-OSC bridge
- 21 Unity components live + integrated
- K11 mini-PC online, three packet streams green
- 73 fast pytest tests passing
- 3D print queue: 31 parts, 7 plates, ~50 hours, ASA matte white + signal orange palettes

The unexpected stack (the moat nobody else holds):
- Mohamed: Calvin Klein model + multi-year N'Ko AI infrastructure work (ASR for Bambara, sigil composer, learning apps) + audit consultancy (Grand Diomande, Claude Partner Network certified)
- LUME wordmark uses N'Ko script — a writing system designed in 1949 by Solomana Kanté for Manding languages (35M speakers). Heritage isn't a feature; it's the spine.
- Wave 9+ roadmap: motion-driven N'Ko inscription on the bar display. The body literally writes the script on the wall in real time.

Market:
Bars, lounges, galleries, retail brand spaces, experiential-marketing agencies. Every venue with a screen-shaped wall and a Saturday-night closing problem. Adjacent: museums, art installs, brand-activation pop-ups, immersive-theater operators. 2-3 individual artists are building toward this aesthetic; none have shipped a productized engine.

Traction:
- 30+ build-in-public videos in production (TikTok @grand.diomande, IG @diomandee, LinkedIn, Substack)
- Substack readership at granddiomande.substack.com (N'Ko + AI-infra posts)
- Modeling audience cross-pollinates discovery
- Grand Diomande consulting brand operating in parallel ($2.5K-$20K tiers, March 2026 launch)

Ask:
Pre-seed / seed funding to (1) ship LUME-001 install, (2) build the productization layer, (3) hire one hardware engineer + one Unity developer, (4) license commercial mocopi-pro pipeline. Use of funds itemized, not "general operations."

Why now:
- Real-time body tracking hit consumer-grade fidelity (mocopi, ARKit) for the first time
- Mini-PCs deliver Unity-grade GPU compute under $1000
- Post-COVID venues need experiences, not screens with images
- N'Ko AI infrastructure hits an inflection — cultural layer becomes a real differentiator

Risks (be honest):
- Sales motion to venues unproven. Mitigation: pilot install + content engine drive inbound; consulting brand provides 6-month buffer.
- Hardware supply chain risk on sensor fulfillment. Mitigation: sensor-agnostic wire format, fallback path already shipping.
- Single-operator bus factor. Mitigation: that's part of what funding solves.

# WHAT TO BUILD

A 10-slide HTML deck (interactive, arrow-key navigation, F for fullscreen, scrollable for backup). Slide order:
1. The Hook (framing-led — for hardware-first variant: lead with the engineering moat)
2. The Problem (venues with closing problems)
3. The Solution (what LUME is)
4. Why Now
5. The Product (depth fits framing)
6. Market & First Buyers
7. Traction & Receipts (use a receipts-grid layout, 4-6 cards)
8. Tech Edge / Moat
9. Team & Why-Mohamed (lean into the unexpected stack)
10. Ask & Use of Funds

Constraints:
- Asymmetric layouts. NOT every slide identical.
- Slide 1: bold-yellow title fills 60-80% of canvas. One subhead line. Caveat cursive accent.
- Slide 7: receipts as cards, JetBrains Mono numerals, condensed labels.
- Slide 9: photo-or-wordmark spot + 3-5 bullets that name the unexpected stack (Calvin Klein × N'Ko AI × hardware install).
- Slide 10: large ask number, itemized use-of-funds.
- Subtle grain texture (SVG turbulence filter).
- Bottom-right slide counter ("03 / 10") in JetBrains Mono.

Output the complete HTML file. Make it gorgeous. Use the asymmetry and overlap rules. Make it feel like a person who builds, not a deck app generated this.
===

---

After variant 1 returns, ask for the other 4

Paste this follow-up:

Now generate 4 more variants in the same style and brand spec, each a separate full HTML deck. Same content, different framing on the opening 3 slides:

V2 — EXPERIENCE-FIRST: Lead with "the room stops being a room. the room becomes a partner." Walk through what a Saturday night feels like in a venue with LUME. Hardware specs are slide 5+, not the headline.

V3 — CULTURAL-FIRST: Lead with the N'Ko / Bambara / unexpected-stack angle. Open with the wordmark on the shell that 35M people read but most VCs haven't seen. Tie to the cognitive-infrastructure thesis. Position LUME as commercial install AND cultural-sovereignty engine. Use the rust accent #C13B1A for slide 1 and 3.

V4 — TRACTION-FIRST: Lead with shipped artifacts. Open with "six months in. 21 components live. K11 deployed. 73 tests green. shell printing. first install ships from this desk." Heavy on receipts, light on vision. Best for VCs who pattern-match on builders over storytellers.

V5 — MARKET-FIRST: Lead with the buyer-side problem. Open with venue economics: "a bar gets 100 bodies through it on a Saturday night. nine of them feel something. the rest scroll. that's the gap LUME closes." Hardware and tech are the moat slide.

Output each as a separate complete HTML deck I can download.

---

When all 5 are done

  • Compare side-by-side with the pipeline-generated variants in `decks/andrewchanvc/*.html` (those will be ready by morning)
  • Pick the strongest. Either the Claude-Design version or the pipeline version, whichever lands harder.
  • Rewrite slides 1, 7, 9 by hand. AI does 90

Why this exists

ab-browser headless can't drive claude.ai (Cloudflare blocks the Playwright fingerprint, and there's no claude.ai vault profile anyway). So instead of automating, I produced this paste-once prompt that condenses everything Claude Design needs in one block. Single paste, full visualization.

Promotion Decision

Attach run IDs, datasets, metrics, and reproduction commands.

Source Anchor

lume-content/vc-pitch/CLAUDE-DESIGN-PROMPT.md

Detected Structure

Method · Evaluation · Architecture