Gym Ops

• 

  Phone QR code for FREELOCITY Software Demo! **Gym Ops is runnable from local host: http://localhost:3000/

Inspiration: Pivoting from Freelocity to Gym Ops

Alston and I met in undergrad when he was my TA for microprocessors. We reconnected as working professionals, and took and inspiration in this hackathon! We both have always enjoyed sports and working out (something not very common in the STEM world). When we heard about the physical health category, we knew we wanted to build SOMETHING... That idea eventually became Freelocity/Velocity Unlimited!

And then, we had to pivot!!! 🫨🫨🫨

We got to a certain point where the app just wasn't demoable. It was a neat and ambitious idea, leveraging computer vision and the accelerator on the phone to provide for free what normally costs hundreds of dollars in specialized gym equipment.

Sooooo... We hard pivoted from Freelocity to Gym Ops!!!

What it does

Inspired by Manufacturing Execution Systems, GymOps uses telemetry and usage-based maintenance to anticipate service, reduce downtime and repair costs, preserve capacity, and protect membership value. It combines OPC UA integration, reservations, incident workflows, advisory ML, and Chunk, a grounded AI assistant.

How we built it

Spec-driven AI development largely with "humans in the loop" for refined specs ovver time.

Endless conversations and planning and then testing, carrying the project over the finish line and working out the kinks.

Challenges we ran into

We were using Replit to build, which was a very sophisticated online IDE. Now Replit has an AI assistant for coding. We are young professionals, but didn't seriously dip our toes into properly using agentic coding until this hackathon. This really opened up our eyes to just how fast you can go from idea to Minimal Viable Product (MVP), but there we're definitely dangers in using it.

Agentic AI and prompt engineering is sooooo good, and so great, which is fine for a hackathon. But the catch-22 is the expertise for the bookends. In spec-driven development, you are basically generating a bible/road map for your assistant to build. The actual coding part of it is rather quick when it can be outsourced to an agent. But now we are doing twice as much work to understand the code we "orchestrated" but didn't write ourselves other than touch-ups here and there.

Accomplishments that we're proud of

While we're gymgoers, we had no clue that was even a thing until exploring specialized weightlifting metrics used by some of the best athletes in the world.

Built a functional MVP.

Building something we can be proud of. (Hopefully) impressed some judges!

What we learned

Learned a ton about velocity-based training Learned about doing Spec-Driven development the RIGHT (and wrong) way

Be a good orchestrator. Drive the spec, but don’t let the spec drive you! Understand the code and purpose. Tokens are not cheap! Don’t utilize agents for every little thing. The ambition of the solution should match the magnitude of the problem!

Additionally, properly research the problem (and solution) from all angles first. That would have saved us from having to restart.

Audio Warning for YouTube submission around 1:30.

...And had fun doing it!

What's next for HealthHackers

Continue doing more hackathons, but we actually planned to contact some local gyms if they would be interested in being clients. We can take this on tour!

Built With

• ai
• docker
• json
• opc
• openai
• react
• replit
• sparkden
• typescript
• uaexpert