Claude Design Nails Wireframes & Decks, Flops on Video
Claude Design's questionnaire acts like a PM for superior wireframes and 90% ready pitch decks, saving hours—but video is only 5/10 and token costs add up fast. Start low-fi to iterate efficiently.
Questionnaire Drives PM-Level Wireframing
Greg Isenberg tests Claude Design (claude.ai/design, research preview) in a live, unscripted workflow for a real product idea: "Senior Brains," a gamified brain exercise app for seniors inspired by Duolingo and Brain Rot app, sourced from ideabrowser.com. Instead of one-shot high-fidelity designs, he starts with low-fidelity wireframes to conserve tokens and refine features.
The standout is Claude's pre-generation questionnaire, which probes like a product manager: primary device (iPhone), mascot tone (gentle, silly, calm), screens to mock (onboarding, daily home, rewards, progress), directions (3), gamification (streaks, XP), accessibility (large text, high contrast, voice controls), exercise types (memory match, word recall), and family caregiver role (visible cheer-ons). Isenberg notes: "Felt like it did a good job at looking at what the idea was and extrapolating from there like a product manager. That was actually quite good."
This yields three distinct low-fi directions: A (warm, card-based, Duolingo-like), B (mascot-forward navigator), C (calendar ritual, less gamey). Each includes full screens with interactions like family hearts or progress journals (e.g., "20% faster word recall"). No tokens wasted on mid-fi—Isenberg rejects it outright: "midfi wireframes are bad... start with lowfi or go hi-fi. You can't be half pregnant."
From chat feedback, he picks Direction A, prompts for hi-fi: "Be a creative director... research Brain Rot and Duolingo... make something the CPO would say 'This is amazing.'" After debug retries (common live errors shown), it delivers clean, usable mockups ready for iteration in ~30 minutes of back-and-forth.
Pitch Decks Deliver Agency-Quality Output
Parallel to wireframes, Isenberg generates a Sequoia-style VC pitch deck for Senior Brains (seed stage, $2M raise, Greg as founder building product/Meta ads). Questionnaire again shines: deck length (5 min), aesthetic (warm/human), style (short topics), clinical vs. consumer (credibility balance).
Result: A near-complete deck (90% nailed with minimal input), covering problem, solution, market, traction placeholders, team, and ask—equivalent to "thousands of dollars of value if you priced the equivalent work from a designer." He calls it the session's highlight, saving hours vs. manual creation. Tradeoff: Assumes MVP exists; real traction data needed for polish.
This fits indie builders or agencies: Isenberg's Late Checkout Agency (latecheckout.agency) serves Fortune 500s like Warner Music/Dropbox with AI products, and he sees Claude accelerating direction exploration (A/B/C like agency pitches) at zero initial cost.
Video Generation Underperforms for Polish
Pushing boundaries, Isenberg requests a 30-second animated ad: mom Ruth and daughter Sarah connecting via app. First output: Social-feed clip (5/10), not cinematic commercial. Iteration for "more cinematic" improves pacing/voiceover but lacks production quality—workable for Instagram Reels/Facebook (seniors' platforms) but not TV-ready.
Limitations surface: Token burn accelerates (chat reports 15-30 min limits), errors require retries/debug, no Figma import tested (future interest for design systems like Apple's). iPad/pencil support speculated for napkin sketches. Overall verdict: Best-in-class wireframes/visuals; video needs work.
Token Management and Real-World Workflow Fit
Core decision: Wireframe-first conserves tokens, sharpens decisions before hi-fi commitment. Tutorials one-shot hi-fi wastefully; live demo exposes stumbles (errors, waits) for authenticity. Isenberg emphasizes: "The only way to know a tool is to get your hands dirty." Potential: $5-15M ARR business via Reels/Facebook, buildable with Claude Code post-design.
Tradeoffs named: Excels solo/indie (fast ideation), weaker teams needing imports/collaboration. Fits product validation: Idea → wireframes → deck → ad prototype in ~1 hour. Future tests: Design systems, Figma integration.
Notable Quotes:
- "I'm blown away by how good these questions are." (On questionnaire; shows PM intelligence beyond basic tools.)
- "The deck alone represents thousands of dollars of value." (Pitch deck output; quantifies time savings vs. hiring.)
- "If I was actually trying to build a business, I would start with the wireframe because that's going to help me figure out what features do I want." (Workflow rationale; prioritizes efficiency.)
- "This gives you that agency feel... as of now, I haven't spent one cent on a token." (Directions A/B/C; democratizes pro output.)
Key Takeaways
- Start every Claude Design project with low-fi wireframes and the questionnaire to mimic PM thinking and save tokens.
- Use specific references (Duolingo, Brain Rot) in hi-fi prompts for familiar-yet-fresh results a CPO would approve.
- Generate pitch decks early—they hit 90% quality fast, ideal for VC or internal buy-in.
- Expect video at social-post level (5/10); iterate but don't rely for pro commercials.
- Run live/unscripted tests: Errors and retries teach more than polished tutorials.
- Pair with Idea Browser for grounded ideas; target Facebook/Reels for senior apps.
- Watch token limits (15-30 min heavy use); parallel projects to multitask.
- Test Figma imports/design systems next for team workflows.