What We Did Today

[This is an ongoing blog series, start your journey here.] With the two-week clock ticking on our core hypothesis—can we rely on vibe design / vibe coding–style tools to ship a viable Farm Stand MVP in just 10 working days?—today’s session zeroed in on collecting outside inspiration fast enough to feed those AI tools tomorrow.

A newly-refined set of five How-might-we (HMW) questions framed everything we did (benefits, onboarding, farmer inventory, shopper discovery, community growth). Pairs had 15 minutes to scour their phones for best-in-class screenshots, drop them into Miro, and tag each image with why it matters for a farm-stand app. A 15-minute share-out followed, exposing patterns around trust, offline maps, and AI-assisted product listings. We wrapped by assigning each teammate a complete farmer-or-shopper flow to prototype this afternoon—feeding those references straight into Google Stitch, Figma Make, ChatGPT, or other AI tools. Monday we’ll smash them together and see if our AI stack is actually sprint-worthy.

Today, we focused on:

• Rewriting sprint goals into five crystal-clear HMW questions
• Collecting & annotating reference screenshots in Miro
• Live critique of external patterns (Google Maps, Airbnb, AllTrails, Carrot, Poshmark, Substack, etc.)
• Handing out end-to-end flow assignments and locking tool choices for the afternoon build-phase

Roles, Tools & Working Styles

A designer-heavy squad—and a few AI-curious generalists—are stress-testing multiple toolchains in parallel, then funneling everything back to Miro for group critique.

Designers

Toolset: Miro Prototype, Figma Make, Google Stitch, Subframe, Lovable, phone-to-Miro uploads
Approach: Visual first, prompt second—every mock must be annotated with a “why,” and AI is treated as a co-designer, not a magic button.

Explorers / Generalists

Toolset: ChatGPT conversational prompting, ad-hoc scripts
Approach: See how far a single chat-driven workflow can scaffold a flow before designers polish.

Key Insights & “Aha” Moments

Even at this early scouting stage, a few themes already point to what could make—or break—the viability of a Google Stitch + Claude Code handoff:

• Maps win hearts. Google Maps/Airbnb tap-to-popup interactions feel like instant MVP material.
• Offline matters. AllTrails’ downloadable areas answer the “no cell signal at the farm gate” problem.
• Trust is currency. Carrot’s neighborhood-verification and Substack’s follow→subscribe model give us low-friction credibility levers.
• AI listing autofill saves farmer time. Poshmark’s photo-to-form flow could be our secret weapon for rapid inventory updates.
• Tool diversity = blind-spot buster. ChatGPT minimalism and Google Stitch hi-fi screens surfaced very different assumptions about scope and polish.

Selected AI Tools and Approaches to Explore

Goal: Prove (or disprove) whether these AI stacks can actually compress design-to-dev enough to hit our two-week MVP target.

Quotes from the Team

“I like the simplicity.” — JP

‍ Context: Affirming that the five HMW questions keep the sprint laser-focused on what will make or break an MVP.

“AI asking you and you giving the answer—letting AI hallucinate—I feel it.” — Magdy

Context: Endorsing a prompt-driven approach where the AI interrogates the designer to co-create screens.

“It’s easier when I have an actual example to show the AI first.” — Priscilla

Context: Explaining why she loads screenshots into Miro’s AI prototype tool before prompting—better context, better output.

Hypotheses We’re Testing Today

We track assumptions like scientists—then poke holes in them tomorrow.

• Map-first discovery will beat list-first in a hyper-local marketplace.
• Screenshot-led prompting yields sharper AI UIs than text-only prompts.
• Running multiple AI tools in parallel will surface better patterns than betting the farm (sorry) on a single stack.
• Trust/verification layers will raise adoption even if they add onboarding steps.

Takeaways for Builders

Early scouting already hints at broader lessons:

• Anchor to proven patterns—Airbnb’s tap-to-popup map is battle-tested.
• Bake in offline support early if your users drive into dead zones.
• Show your AI examples; good prompts start with visuals.
• Don’t marry one AI tool too soon; diversity exposes blind spots.
• Trust features (reviews, verification) are non-negotiable in community marketplaces.

What’s Next

Coming up:

• Afternoon sprint: everyone builds a full farmer- or shopper-flow in their chosen AI tool.
• Toolsets: Miro Prototype, Figma Make, Google Stitch, ChatGPT, Lovable, Subframe.
• Documentation: daily Notion journals (template inbound), screen captures, Monday live walkthrough.

Final reminder: This isn’t about perfection—it’s about proof. We’re building—and breaking—in public to answer one question: Can vibe-style design & coding actually ship a usable mobile farm-stand MVP in two weeks?

‍

‍

‍