Focus of the Workshop

This isn’t just a sprint. It’s a public experiment.

We kicked off a two-week sprint to explore a bold question:

Can AI tools—like Google Stitch and Claude Code—really help a small, cross-functional team build a production-worthy app faster, cheaper, and better than traditional workflows?

At stake isn’t just one app. It’s the future of product delivery. Agencies, entrepreneurs, and in-house teams are all under pressure: AI hype is high, the fear of being left behind is real, and clients increasingly ask, “Can we do this faster with AI?”

So we’re stepping into the ring—hands partially tied behind our backs—to test how far AI-first workflows can take us.

We’re not here to promise perfection. In fact, if this fails, that’s part of the point. This is about documenting reality. The stumbles. The breakthroughs. The workarounds. The human messiness of building in a new medium.

Our sprint goals stack like this:

1. An awareness campaign showing how we really work with AI—not just talk about it.
2. A tooling testbed to help us (and clients) assess if “vibe coding” is hype or legit.
3. A real-time, unvarnished case study for our agency and the industry.
4. A repeatable playbook we can refine and rerun in 6 months.
5. And maybe—just maybe—a real product that solves a real problem for small farmers.

What We Did Today

We used Day One to align the team, define our purpose, lay out the schedule, and establish the parameters of the experiment. Most importantly, we grounded everything in user value by mapping early journeys for both farmers and shoppers.

• Aligned on sprint goals, timeline, and communication rhythm
• Selected and committed to AI tools (with flexibility for experimentation)
• Began mapping user journeys for farmers and shoppers
• Identified content types and cadence: Notion journals, roundtables, video updates
• Captured shared understanding around the purpose, priorities, and expected outcomes

“If I want to build fast and securely, I need a tool that doesn’t make assumptions for me.”

Tools in Play

Everyone picked their starting weapon. Day One surfaced strong opinions on AI tooling—particularly around portability, code audibility, and design-developer handoff. The goal isn’t perfection—it’s breadth. We’re testing a spectrum of tools to see what sticks and what breaks.

Team MemberTools and Notes

Developer - Tools: Claude Code Notes: Prioritized developer-grade output, security auditability, and freedom to escape tool lock-in. Clear skepticism of tools like Lovable and Bolt that embed Supabase keys in client code.

Designers - Tools: Figma Make, Miro AI, Hand Sketch + AI Interpretation Notes: Tools will vary, but all output must be visual (screenshot, video, or sketch). Encouraged to use AI as a co-creator, even from hand-drawn sketches.

Explorers - Tools:  Bolt, Lovable, Google Stitch Notes: Secondary tools added for breadth. Non-dev teammates will explore their potential for fast UI gen and concept modeling.

Key Insights & “Aha” Moments

Even in Day One, patterns are starting to emerge. While we haven’t written a line of final code or drawn a polished screen, we’re already spotting tension points between usability, portability, and dev trust. This is why we’re doing this—to learn these things before recommending them to clients.

• Tool Lock-In is Real: Platforms like Lovable and Bolt embed decisions you can’t easily undo—especially when it comes to backend configuration like Supabase keys.
• AI ≠ Magic. It’s Just Fast: Tools like Claude Code and Figma Make still rely on strong inputs. There’s no “auto solve” button—yet.
• Early Output Creates Real Discussion: The ability to visualize concepts early (even roughly) allowed developers to evaluate feasibility in parallel, not sequentially.
• AI Doesn’t Eliminate Process—It Reframes It: We’re still mapping flows, setting constraints, and doing user journeys. The difference? It’s happening faster, with tooling guiding some of the branching paths.

Quotes from the Team

We’re capturing real-time reactions, both philosophical and technical. These quotes reflect the tension, curiosity, and cautious optimism running through the team.

“This isn’t about getting it right. This is about finding out where it breaks.”

“Let’s stop pretending this is all magic. Our job is to see if the wizard behind the curtain is real—or just good marketing.”

“Even if this fails completely, the content and insights we’ll share are still valuable—and that’s the point.”

Hypotheses We’re Testing

We’re treating this like a scientific sprint. That means writing down what we think will happen—then checking back in later to see what’s true, what broke, and what surprised us.

1. AI-first prototyping can meaningfully compress the design-to-dev workflow
2. AI-generated code is only as good as its deployment context (security, structure, auditability)
3. Most no-code tools introduce long-term constraints that outweigh short-term speed
4. You can run a public experiment and still ship something useful
5. Clients will respond better to real-world case studies than generalized AI hype

Takeaways for Builders

Here’s what other product teams, developers, and design leads might learn from our Day One:

• Choose tools with a clear exit plan—you might outgrow them fast
• Auditability matters, especially when AI is generating code
• Prototyping is faster, but still needs structure
• Parallel thinking is powerful—devs working alongside sketchers = better outcomes
• Process still matters. You still need a plan—even if AI is holding the pen.

What’s Next

We’re entering the sketching and concept exploration phase. Each team member will bring flows to life using their chosen tools, feeding them into our shared design vault. Meanwhile, daily journals begin flowing into Notion for synthesis and public updates.

Here’s what’s on deck:

• Day 2: Compare & Sketch with multiple toolchains
• Journals: Daily reflections begin (typed or recorded)
• Farmer Outreach: Start lining up test participants