How The Bounce Generation turned a Lovable-generated prototype into a launch night app with zero downtime
We rebuilt a Lovable prototype into the app that ran entry, orders and help at a 63-guest launch party.
The prototype looked ready until we tried to run a party on it
The Bounce Generation was launching with The First Gathering, an invite-only night in Gurgaon. The founder had built the guest app in Lovable. The screens were all there: a home page, a menu, a programme and a check-in flow. It looked close to done.
Underneath, it would not hold up in front of a real crowd. A hardcoded phone number gave anyone who knew it full owner access. Payments were simulated, so no order was ever really confirmed. Guest lookup ran through Lovable's own connector, nothing limited how often a form could be hit, and two guests ordering the last item at once could push stock below zero.
The app also had no picture of how the night would actually run. The kitchen and bar had no screens of their own. Staff had no way to confirm a payment. A guest who needed help had no way to reach a specific person.
And the party was five days away.
A rebuild shaped around how the night would actually run
We kept the founder's design and wording and rebuilt everything underneath. Every feature was tied to a real moment on the night: a guest at the door, an order at the bar, a request for help. Every flow also got a paper fallback in case a phone or the WiFi let someone down.
We moved the app off Lovable and onto infrastructure the founder owns: a Supabase database in Mumbai and Cloudflare Workers. We removed the hardcoded owner backdoor and gave each staff member a one-time login link. Every table was locked down with row-level security, forms got rate limits, and stock updates became atomic so an item could never sell past zero.
The back door was closed before a single guest logged in.
Entry became approval-based. Guests requested access, the founder approved each one, and a guest's phone number became their login. At the door, staff checked guests in by QR code or by name search, with a printed list ready as backup.
Only approved guests could get into the app or the venue.
We built ordering around a real bar. Guests ordered food and drinks on their phones and paid at the bar by UPI. Staff confirmed each payment before anything was made. Orders went straight to the kitchen or bar screen and moved from New to Preparing to Completed.
Every order was confirmed by a person before it was prepared.
Every lead got their own view. We built seven role-based views: Guest, Studio for the host, House, Kitchen, Bar, Experience and App Support. Each help request was owned by one lead at a time and could be escalated to the host. The founder could edit the programme live from Studio.
Each person saw only what they needed to act on.
We put the night on the walls and rehearsed it. The venue TV ran a live board showing what was on now and next, plus the arcade Score to Beat, which updated itself from submitted scores. Slack alerts kept the core team in the loop. Before doors opened, we ran every flow on real phones.
Nothing on the night was being tried for the first time.
What did the rebuild actually deliver on launch night?
5 days from prototype to a live event.
The founder's Lovable prototype became a secured app running on the founder's own infrastructure in five days, with time left to rehearse every flow on real phones before doors opened.
63 guests checked in through the app.
About 70 guests were approved in advance. 63 arrived and were checked in by staff using a QR code or a name search, with a printed list on standby.
82 orders went from phone to kitchen and bar.
Every order was paid at the bar by UPI, confirmed by staff and sent to the right station. The kitchen and bar each worked from their own screen all night.
16 help requests raised, 16 resolved.
Each request went to one named lead who owned it until it was closed. Guests always knew someone specific was on it.
Zero downtime with 40 people on the app at once.
At the busiest point of the night, around 40 guests and staff were using the app together. It stayed up from the first check-in to the last order.
We had a prototype that looked finished and a launch party five days away. Once I sat down with Shivam, I realized how much more work it needed before it could actually run on the night. Shivam and his team at Agent Loopr found problems I didn't even know we had, fixed them, and built the parts we hadn't thought about, like how the bar confirms a payment. On the night, everyone was talking about the app, and they loved using it.
Tools and technologies
used on this project
Frontend
The guest app, staff views and the venue TV board, rendered on the server so pages load fast on mobile data.
Backend & Data
Database, logins, access rules and live updates, hosted in Mumbai close to the guests.
Ops & Build
Team alerts during the night and the AI coding tool used for the rebuild.
Started as a Lovable prototype, now running on infrastructure the founder owns.