Aakash GuptaWe prototyped 5 features in 84 mins (Bolt, Cursor, Lovable, Replit, v0)
CHAPTERS
Why AI prototyping compresses feedback cycles from months to hours
Colin frames AI prototyping as a way to test many more solutions quickly by getting ideas into a realistic, in-product flow. The goal isn’t shipping more features faster, but learning faster from customers with higher-fidelity experiments.
Bolt live build: paid booking links for Apollo’s scheduler (setup + UI grounding)
Aakash proposes a concrete feature: paid booking links inside Apollo’s meeting scheduler. Colin starts from zero in Bolt, grounding the prototype in Apollo’s existing UI via screenshots for near-1:1 styling.
Using AI-generated PRDs as the backbone for iterative builds
Instead of prompting for code first, Colin prompts Bolt to create a plan—resulting in a PRD that becomes a persistent reference. This reduces context loss over long chats and enables incremental development against phases.
Iteration tactics: reflection, small chunks, and avoiding painful undo
Colin explains the operating model: break work into small requests, ask for reflection, and sometimes ask for no-code previews before implementing. This improves control and reduces the cost of undoing large, bundled changes.
Admin workflow prototype: edit page + pricing toggle and persistence
The prototype expands from list view to an editable detail page where pricing can be enabled and set. The workflow saves the price and surfaces it back on the admin overview to confirm end-to-end behavior.
From idea to testable prototype: preview page + payment collection
Colin adds an end-user preview/booking page and then layers in time selection and payment info. A small bug emerges (state loss across tabs), illustrating practical debugging and iteration in AI-built prototypes.
AI design speed vs Figma and why code prototypes enable richer research
They compare AI prototyping to Figma: even strong Figma users may be slower for multi-step flows, and PMs often aren’t Figma-proficient. Code prototypes also allow analytics and real interactions (especially for AI features) that Figma can’t replicate well.
Tool landscape and choosing the right stack (chatbots vs web tools vs IDEs)
Colin categorizes the market into three groups: chatbots, web-based builders, and developer IDEs. He explains when databases/servers matter and why Replit is strongest for full-stack prototypes, while Bolt/Lovable/v0 excel for fast client-side work.
Cursor tutorial: moving from Bolt to an IDE and building AI sequences
Colin exports the prototype into Cursor, runs it locally, and uses Cursor’s Composer to build a new Apollo-like flow: ‘Create sequence with AI’. A misstep (missing context files) demonstrates Cursor’s power and its need for correct context management.
Cursor result: multi-step questionnaire + social proof for AI-generated sequences
With proper context, Cursor implements a pre-generation Q&A flow and then outputs a generated sequence augmented with social proof. The demo highlights how prototypes can replace long PRDs by letting stakeholders ‘experience’ the idea directly.
Lovable tutorial: Figma-to-code via Builder.io (and limits of screenshot-only)
They test Lovable’s conversion pipeline using Builder.io’s Figma plugin. A screenshot-based attempt errors, then a proper layered Figma design converts successfully, showing why structured design metadata produces closer 1:1 results.
Lovable cleanup: element selection, targeted edits, and when to stop polishing
Colin demonstrates Lovable’s element selector to remove an unwanted navigation artifact. They note that for prototyping, minor visual imperfections often aren’t worth extended iteration time.
Replit tutorial: full-stack to-do app with database, auth, and deployment
Replit is shown as the ‘full-stack’ step: generate a React to-do app, add a database, persist items, then add authentication with a simple prompt. They also cover deployment options and the tradeoffs of agent automation (including accidental data loss).
v0 tutorial + closing takeaways: community components and the prototyping revolution
Colin positions v0 alongside Bolt/Lovable as a fast UI prototyping tool, with a unique advantage: reusable community components and Vercel deployment. The episode closes by reinforcing that AI prototyping is primarily a communication and learning accelerator, not a shortcut to production code.
