How a Data Scientist Used CreateOS to Build a Travel Gym Tracker from a Single Conversation

• —Gym finder with real-time crowd alerts built and deployed in a single session at the AI Impact Summit, Delhi
• —Mobile and desktop UI generated simultaneously, ready for use across devices
• —500+ lines of code produced through natural language configuration, with no manual coding
• —Deployed to a live URL via GitHub within the session, actively in use by the builder

Ankit is a Data Scientist at HBO Max and a self-described converted vibe coder. He came to natural language development as a programmer with reservations, initially sceptical that building through plain language could meet the standards of someone who writes code professionally. By the time of this session, he had used most of the major tools in the space and arrived with a clear, specific idea and a willingness to see how far it could go in a single conversation.

Staying consistent with gym training while travelling is a problem with several interlocking parts: finding a gym near the hotel, knowing whether that gym has reliable Wi-Fi, understanding how busy it is before making the trip, and building enough accountability into the routine to actually go. No single tool addresses all of these at once, and for someone who moves frequently and trains regularly, the friction compounds quickly.

Ankit wanted an application that could surface nearby gyms in any city, flag crowd levels in real time using traffic data, and keep him accountable through notifications tied to his workout activity. The use case was practical and personal: a tool built around his specific travel and training habits, not a generic fitness app configured for an average user.

The business need was specific: a personalised fitness and travel tool, built to one person's exact requirements, deployable immediately and usable anywhere. CreateOS solved the build problem by taking the idea in plain language and translating it directly into a working, deployable product, with AI reasoning through the underlying infrastructure and generating a first version without a single line of code written by hand.

The output included a fully responsive interface with both mobile and desktop layouts generated simultaneously, a GitHub deployment button available from the first preview, and over 500 lines of functional code produced through natural language configuration alone. Once satisfied, the application was deployed through GitHub under the name AnyWhereGYM, producing a live URL the builder is actively using.