We Built a Working App in 30 Minutes, Live, Twice, at Boston Tech Week
At #BosTechWeek we sat down in a room full of curious people with a blank page and a 30-minute timer, let the room vote on what to build, and shipped a real, deployed web app before the clock ran out. Here's exactly how, and why it's how we prototype client work now.
The nerve-wracking part: no slides, no canned demo, no safety net. Just the two of us, a product that didn't exist yet, and an audience that wanted to see how this actually works, founders, builders, operators, and a few skeptics. They picked the idea. We designed it, coded it, generated the imagery, and deployed it for real while everyone watched. No finished work to point at, just the work itself.
We loved it so much we did it twice.
Two rooms, two brand-new apps
We ran the same live-build session at two Boston Tech Week events:
- Wednesday, May 27 in the Hogsmeade Room at CIC Cambridge (245 Main St)
- Thursday, May 28 at Venture Café Cambridge, a cozier room that was part of a much larger Tech Week event (thank you to everyone who packed in)
The format was simple and a little chaotic on purpose: Alec drove design, Mark drove engineering, and the audience voted on which of three AI-generated ideas we'd actually build. Then we started a 30-minute timer and went.
"Right now this is a blank page. By the end of this session it'll be a working product, built from scratch, in real time."
What came out the other side:
Wednesday → "Brand Intel Lookup." Type in any consumer brand and get three instant cards: is it currently being sued, where is it putting its money lately (marketing, hiring, expansion), and, if it goes on or in your body, what's actually in it. A 20-minute research slog across PACER, news, and ingredient labels, collapsed into one search box. Ten seconds, one search. Try it live →
Thursday → "Boston Like a Local." Pick a neighborhood and a vibe and get five or six genuinely non-touristy recommendations: the spots locals actually love, with the tourist trap each one replaces. (Sorry, Faneuil Hall.) Skip the tourist trail. Eat where Boston eats. Try it live →
Both went from idea to a live, deployed web app before the timer hit zero.
How you actually build a real app in 30 minutes
People assume "live coding" means typing fast. It's the opposite. The speed comes from the process, not the keystrokes. Here's the actual loop we ran in the room:
- Idea to PRD. We fed the winning idea into a Claude project pre-loaded with our PRD instructions. In about three minutes we had the problem, the users, the scope, and the data shape written down.
- PRD to code. That spec went straight into Claude Code, which scaffolded the app while we narrated what was happening.
- Design in parallel. While the app took shape, Alec shaped the look and feel (buttons, type, layout) then handed it off with a commit.
- AI imagery, compared live. We ran the same image prompt through Gemini, ChatGPT, and Higgsfield side by side and picked winners on the spot.
- Polish and deploy. Mobile-first tweaks, then a live deploy to Vercel.
- Show the receipts. We left the Claude transcript, the GitHub commits, and the Vercel deploys up so nobody had to take our word for it.
The crowd favorite, though, had nothing to do with the code. Every time we stopped typing and just talked to Claude Code out loud, dictating a full prompt with Wispr Flow and watching a clean, punctuated paragraph of instructions land on screen in about a second, the room laughed in awe. Nobody expected the words to keep up that well. By a mile, it was the highlight of both demos.
Why this matters (beyond the demo)
This isn't a party trick. It's exactly how we now prototype and validate real client work. We can put a believable, clickable, deployed version of an idea in front of stakeholders in hours instead of weeks. The 30-minute timer just made the point impossible to argue with.
"I was amazed to see a working POC of an idea from the audience come together live."
A senior engineering manager who was in the room
A few people asked afterward whether we had a write-up they could pass along to their teams. This is it.
The process is the product. A tight PRD, a model that can scaffold from it, a designer and engineer working in parallel, and a willingness to ship something imperfect and iterate. That's the whole game. Everything else is narration.
Stick around: Matt McGuire on which features actually made money
Right after our Thursday session at Venture Café, our contract Product Manager Matt McGuire (the PM we bring in on product work) ran his own workshop, and it's the perfect companion to a "build it fast" talk: once you've shipped, how do you know what was worth shipping?
His session, "You Shipped All Quarter. Which Features Made Money?", walked the room through his product, Vektis, and the question it's built to answer for B2B SaaS teams: connect your GitHub and CRM, and you can finally see which features actually drive revenue and which quietly correlate with churn.
Building fast is only half the loop. Knowing which of those builds earned their keep is the other half, and it's why we like working with people like Matt.
And yes, there was swag
We may have gotten carried away. Stickers and a sign that sums us up pretty well: "We're a designer and engineer dad duo. After the kids are asleep, we geek out about AI and love sharing what we're learning."
