Claude Design Masters Wireframes & Decks, Flops on Video
Claude Design delivers agency-level wireframes via smart PM-like questions and 90% solid pitch decks from minimal input, but video is only 5/10—prioritize low-fi wireframes first to save tokens and refine ideas.
Why Wireframes First Saves Time and Tokens
Greg Isenberg tests Claude Design (claude.ai/design, research preview) on a real workflow: turning a gamified brain-training app idea for seniors—'Senior Brains,' pulled from ideabrowser.com—into wireframes, hi-fi mocks, a VC pitch deck, and a 30-second video ad. He rejects one-shot high-fidelity designs, insisting on low-fi wireframes to constrain features, avoid token waste, and mimic agency processes. 'Why do I want to create a wireframe first? Because I don't want to waste tokens... it's going to help me figure out what features do I want.' This decision stems from experience with LLM tutorials that gloss over costs and iterations; low-fi forces sharp product decisions before visual polish.
He grounds the tool with a screenshot of the idea brief, inspired by Duolingo's gamification and Brain Rot's mascot chaos, but toned for seniors: gentle, silly, calm. Primary device: iPhone. Screens: onboarding, daily home, session, rewards/progress, snacks. Three directions at lowest fidelity. Gamification: streaks, XP. Accessibility: large text, high contrast, voice narration, simplified toggle. Family caregiver role: visible cheers, not prominent.
Questionnaire Thinks Like a Product Manager
Claude's pre-generation questionnaire extrapolates deeply, asking non-obvious questions like 'How prominent is the family caregiver in the main app?'—picking up subtle idea nuances without explicit prompting. "I'm blown away by how good these questions are... Felt like it did a good job at looking at what the idea was and extrapolating from there like a product manager." This PM simulation generates three distinct low-fi directions:
- Direction A (Warm Stack): Card-based home with clear action, small mascot sidekick, Duolingo-adjacent calm.
- Direction B (Mascot Forward): Chatbot-style navigator (Bean mascot) cheering like family/livestream likes.
- Direction C (Calendar Ritual): Habit-focused scrollable path, crossword vibe, progression feel.
Chat votes Direction A. Each includes full screens (onboarding: 'Hello, I'm Bean'), session results (e.g., '11/12 memory match, 20% faster word recall'), and progress journals. Outputs mimic agency pitches with stories per direction, zero cost yet. Napkin sketch tool for freehand (pencil top-right post-build) noted for iPad potential.
Hi-Fi Iterations and Pitch Deck Gold
Prompting hi-fi on Direction A: "Be a creative director... research Brain Rot and Duolingo... something the CPO would say 'This is amazing.'" Initial errors (auto-retries, debug info) highlight live realities vs. polished tutorials—"This is why I'm doing this live stream... you don't put the errors in." Succeeds with clean, usable mocks ready for 30 minutes of back-and-forth.
Pitch deck for $2M Sequoia raise: 90% nailed with minimal input (5-min pitch, seed stage, Greg as founder, warm/human/Sequoia aesthetic, consumer-credible). Thousands in designer value; auto-scripts, short topics. Represents core strength: rapid, high-quality assets from briefs.
Tradeoffs surface: Mid-fi wastes time ('can't be half pregnant'), token burn (15-30 mins per X reports), no Figma import tested (future design systems intrigue, e.g., Apple recreation).
Video Falls Short of Commercial Polish
30-second animated ad: Mom Ruth and daughter Sarah connecting via app. First: social-feed clip (5/10). Cinematic reprompt improves but lacks TV-commercial depth—workable for posts, not pro. Weakest link; stick to static strengths.
Workflow Fits Indie Builders and Agencies
Verdict: Best-in-class wireframes/visuals; pitch decks save hours; video mediocre. For solo founders/indies: Idea → low-fi wireframes → pick direction → hi-fi/deck → handoff to code (Claude Code). Agencies: Rapid directions for clients (Warner Music/Dropbox via his Late Checkout Agency). Conserve tokens, embrace errors/iterations. Potential $5-15M ARR business from Senior Brains via FB ads/Reels.
"Claw design is a best-in-class product for wireframes, visual designs, not so much videos. You'll see why by the end."
"The deck alone represents thousands of dollars of value if you priced the equivalent work from a designer."
"Mid-fi wireframes are bad. You want to start with low-fi or go hi-fi."
"The only way to know a tool is to get your hands dirty."
Key Takeaways
- Start every Claude Design project with low-fi wireframes to refine features and save tokens—avoid one-shot hi-fi.
- Leverage the questionnaire for PM-level extrapolation; answer thoroughly for grounded outputs.
- Generate 3 directions to mimic agency pitches and force conceptual range.
- Pitch decks hit 90% quality from minimal prompts—ideal for VC/ideas validation.
- Skip video for now (5/10); use for social clips only.
- Handle errors live: auto-retries/debug; refresh if stuck.
- Ground with screenshots/briefs (e.g., ideabrowser.com) for better context.
- Post-build: Pencil tool for sketches; test iPad for freehand.
- Scale to businesses: $5-15M ARR potential from validated designs + targeted ads.