Aakash GuptaWe prototyped 5 features in 84 mins (Bolt, Cursor, Lovable, Replit, v0)
At a glance
WHAT IT’S REALLY ABOUT
AI prototyping demos: build and test product features faster
- They build a paid booking-links prototype for an Apollo-like scheduler in ~10–15 minutes using Bolt, starting from screenshots and iterating via small, natural-language prompts.
- They show how AI-generated PRDs and “plan/reflection before code” prompting improve iteration quality and help maintain context over longer projects.
- They export a prototype into Cursor to demonstrate IDE-based AI building and debugging, highlighting the importance of providing correct file context and the ability to switch models for harder problems.
- They demonstrate Figma-to-prototype workflows using Lovable + Builder.io, noting higher fidelity when converting real Figma layers versus a flat screenshot, plus element-level selection for targeted edits.
- They use Replit to add full-stack capabilities (database, auth, deployment), explaining when to move beyond client-only prototypes and the tradeoffs of agentic automation and complexity.
IDEAS WORTH REMEMBERING
5 ideasSpeed matters most when it compresses the learn-test loop, not just delivery timelines.
They contrast old cycles (months to ship an experiment) with AI prototyping that can produce a testable flow in hours or days, enabling many more customer conversations and hypothesis tests.
Start prototypes with a plan/PRD to reduce drift and re-explaining context.
Colin configures Bolt to generate a PRD from a “plan” prompt, then iterates by referencing phases, which helps when chat context degrades in longer sessions.
Use “plan first” and “reflection” prompts to improve output quality and control changes.
They repeatedly ask “How would you do this? Don’t write code” to preview approach, then implement; reflection helps the tool compare to screenshots and tighten design fidelity before adding features.
Keep prompts small: build the minimum page/flow, then layer in features.
They recommend creating the simplest working version (e.g., preview page), then adding time selection and payment UI, because big multi-feature requests increase failure risk and rework.
Debugging is often about product state and environment, not just code bugs.
A missing price display was traced to opening preview in a new tab (state not shared); fixing navigation behavior (same tab) restored expected UI without major refactors.
WORDS WORTH SAVING
5 quotesI think really what it's about is just getting through more solutions faster.
— Colin Mathews
I think of, of AI coding tools like this one or AI prototyping tools as tools for communication primarily in this context.
— Colin Mathews
The faster you can build doesn't mean that you should just ship more features, like ship 10, 10 times the features or 100 features, uh, in the time that you could have shipped, you know, one, but instead that you should have 10 more or 100 more co- conversations with your customers, and like richer and deeper conversations with your customers, where they can b- provide you with better feedback r- rather than just one conversation with your customers.
— Colin Mathews
It's better to have like lo- like I said, lots of small changes rather than trying to get the perfect prompt up front.
— Colin Mathews
For whatever reason, I've never been able to figure out what image sizes Substack wants from me. Like, every time I put something in the preview image, it seems to get cut off in some weird way.
— Colin Mathews
High quality AI-generated summary created from speaker-labeled transcript.
