Nymbl AI Hackathon Recap:
From Idea to Production
Thirty applications in one day, built by the people who will use them.
Last month, we filled a room at the Aloft Seaport in Boston with clients, prospects and our own Nymbl team. Everyone arrived with an idea, and nobody arrived with a single line of code. By the end of the day, thirty teams had designed and built working applications in Nymbl Studio, then stood in front of the entire group and showed them off.
In our post-event survey, we asked attendees how long they would normally expect that kind of work to take. Most said weeks, and a couple said months. Every one of them finished it the same day, and watching those demos is a moment our team will be thinking about for a long time.
The hackathon was never meant to be a contest of clever prompts. We designed it as a live test of the same framework our Nymbl team uses every day to take client ideas directly into production, running at full speed, with our clients in the driver's seat. Here is how the day went.
How the Day Unfolded
Design
Business problems become epics, user stories and acceptance criteria in Nymbl Helix.
Build
AI coding agents in Studio build against the plan while each team steers in a live preview.
We split the day into two halves, with the morning session being all about design. Each team took a real business problem and, with our architects working alongside them, turned it into a plan made of epics, user stories and acceptance criteria, all captured in Nymbl Helix. It is the least glamorous part of building software, but it is also the part that decides whether what you build actually works.
After lunch, the energy in the room shifted. AI coding agents in Studio began building against those plans, and each team watched its application take shape in a live preview, testing it, steering it and asking for changes as they went. Our team moved from table to table all day, with short lessons to set up each stage, and by late afternoon every team was ready to present.
When we added it all up, the numbers even surprised our team. In a single day, those thirty teams wrote 501 epics, 1,356 user stories and 4,365 acceptance criteria, and worked through 1,202 prompts between design and build. That comes to roughly 45 user stories per application, each with criteria a team could test against. That depth is what separates a working application from a quick prototype, and it is exactly what we ask of every project we take on.
One day, by the numbers
30
501
1,356
4,365
1,202
What Got Built
The range of projects says a lot. Teams built a grant application portal, a biotech competitor tracker, a QA/QC scheduler, a sales demand forecast, a software pricing quote tool, an EDI file explainer and a client invoice portal. Some built for fun, too. A backyard chicken app even made the list.
Three teams took the podium.
01
Stadium Parking Analytics
Turns an overhead camera into a live picture of a stadium lot. Every car is tracked, every space is marked taken or free, and the exit queue is measured lot by lot: cars out per minute, peak queue and time to leave.
01
56
162
02
Repo Visualizer
Reads an uploaded codebase and draws it as one map, so a team can see how the logic fits together. Pick a flow and each step lights up the functions it touches.
1,433
2,190
245
03
ShelfIQ
Turns an overhead camera into a live picture of a stadium lot. Every car is tracked, every space is marked taken or free, and the exit queue is measured lot by lot: cars out per minute, peak queue and time to leave.
15
57
163
None of these are slideware. Each one does real work on real inputs, and each one was built in an afternoon.
Why it Worked: The Framework, not the AI Hype
Speed with AI is easy to promise. Speed that actually holds up in production is harder. The hackathon worked because every team followed the same governed path we use on client work each and every day:
01
Requirement. Start with the business problem, written as epics, stories and acceptance criteria.
02
Spec and plan. A person approves both before any code is written.
03
Build. AI agents build each change in its own isolated workspace.
04
Preview. The team sees the working application live and steers it.
05
Release. The change is reviewed, merged and shipped, with every line traced back to the requirement it serves.
The AI moves fast, but the judgment stays with people. Every change is checked for code quality, standards and vulnerabilities before a person reviews it. The code is real, standard and yours, with no proprietary platform and no lock-in.
That is why a one-day event is a fair proof point. The hackathon did not use a shortcut version of our process. It used the process itself, compressed into a day. The path from idea to working application is the same path that carries it into production.
In their words
"Amazing vibes, the Nymbl team was incredibly supportive throughout the day and the short lessons set everyone up for success. Got to experience vibe-coding within an environment that made it intentional and structured."
"What an amazing product that you're building. I can't wait to see how we integrate [it] within Sarepta!"
"Fun, fast paced event where you'll learn from some of the most experienced app/system architects around."
"Working with AI is the future. Learning how to do that effectively and efficiently is going to be the difference between being replaced by AI or not."